While these things have stood the test of time, AI has introduced a feast of new ingredients. On October 15th, we are running the first cohort of the Reforge AI Growth course. The course is four weeks, ≈500 pages, 20 hours of videos, and weekly sessions with featured guests to deconstruct growth in AI. I created it with Estrella (Head of Marketing at Clay), (Head of New Products at Notion), and it’s instructed by (Advisor and former Chime, Yelp). Details below.

Enroll In AI Growth Course

Or view all our courses such as AI Foundations, AI Productivity, AI Strategy, AI Leadership.

Reforge AI Growth

We are in the most strategically intense environment in tech history. AI isn’t just another technology shift, it’s compressing years of competitive evolution into months. Companies that dominated their markets for decades are watching as AI-powered alternatives capture new markets in weeks. Meanwhile, unknown startups are scaling from zero to millions in revenue faster than ever before.

This course is your strategic guide for navigating AI-driven growth transformation.

Unlike a list of tactics that expire overnight, this course provides the strategic frameworks and mental models you need to create your own tactics as the rules of growth are being rewritten in real-time. You’ll learn how to adapt proven growth methodologies for an AI-transformed landscape where customer expectations spike overnight, traditional defensibility crumbles, and entirely new growth mechanics emerge.

Created By Leading Operators

In addition, contributions from , Patrick Campbell, Ethan Smith (Graphite), and (Rippling, ).

Who This Course Is For

Primary Audience:

• Growth Leaders orchestrating acquisition, retention, and monetization strategies

• Founders & CEOs making strategic bets on AI-enhanced growth models

• Product Managers building AI-native or AI-enhanced products

• Marketing Leaders adapting go-to-market strategies for AI disruption

• Strategy Professionals evaluating competitive dynamics in AI markets

Prerequisites:

• Working knowledge of growth fundamentals (funnels, loops, metrics)

• Experience with product-market fit and growth model development

• Familiarity with SaaS business models and unit economics

• No technical AI expertise required

Enroll Here

Course Curriculum

The AI Growth Imperative

Understanding the Strategic Landscape

Why Growth Matters More Than Ever

• Lesson 1.1: The Three Advantages of Growth

  • Learn how defensibility, resources, and learning compound exponentially in AI markets, creating unprecedented advantages for fast-moving companies.

  • Understand why AI makes traditional competitive moats more fragile by reducing switching costs, enabling rapid feature replication, and democratizing capabilities that once required years of development.

  • Explore the acceleration of winner-take-all dynamics where the gap between first and second place widens faster than in any previous technology shift.

• Lesson 1.2: The Compressed Timeline Reality

  • Discover how competitive cycles have compressed from years to months, requiring new frameworks for strategic decision-making under uncertainty.

  • Analyze case studies of Stack Overflow’s traffic collapse, Chegg’s sudden obsolescence, and other examples of instantaneous product-market fit collapse.

  • Master the art of making decisive strategic moves without perfect information, understanding why waiting for complete data guarantees competitive failure.

The Evolving Growth Equation

• Lesson 2.1: Timeless Principles in a Transformed World

  • Master the fundamental growth equation (Acquisition + Retention + Monetization) × Defensibility and understand which components AI amplifies versus transforms.

  • Identify what changes in growth strategy (tactics, channels, velocity) versus what endures (systematic thinking, compound effects, value creation).

  • Build growth strategies that are systematic, deterministic, sustainable, and repeatable even as the tactical landscape shifts weekly.

• Lesson 2.2: The Four Fits Framework

  • Evaluate Problem-Solution Fit when AI expands the solution space and enables previously impossible use cases.

  • Navigate Audience-Solution Fit as AI democratizes skills and expands your addressable market beyond traditional personas.

  • Adapt Distribution-Solution Fit as answer engines replace search engines and new discovery mechanisms emerge.

  • Restructure Business Model-Solution Fit to account for variable AI costs, usage-based expectations, and outcome-oriented pricing.

Assessing Your AI Risk Profile

• Lesson 3.1: Finding Your Foundation

  • Apply the Growth Model Canvas specifically designed for AI products to map your current position and identify strategic opportunities.

  • Understand your starting position across acquisition, retention, monetization, and defensibility in the context of AI transformation.

  • Identify the highest-impact transformation opportunities that align with your company’s strengths and market position.

• Lesson 3.2: The 18-Factor Risk Assessment

  • Use Case Factors: Assess your vulnerability based on automation potential, customer tech-forwardness, usage frequency, and relationship importance.

  • Growth Model Factors: Evaluate channel stability, growth loop integrity, and customer relationship dynamics in an AI-disrupted landscape.

  • Defensibility Factors: Analyze your proprietary data advantages, network effect resilience, and switching cost durability against AI competitors.

  • Business Model Factors: Examine pricing model vulnerability and unit economics pressure from AI-driven market changes.

  • Develop a comprehensive scoring methodology and create an action plan prioritized by risk and opportunity.

Acquisition in the AI Era

Building Compounding Growth Systems

Growth Loops Reimagined

• Lesson 1.1: The Evolution of Growth Loops

  • Understand why growth loops create compounding returns while funnels produce diminishing returns, and how this difference becomes even more critical in AI markets.

  • Master the three core categories of growth loops (Viral, Content, and Paid) and identify which types align with your product’s natural mechanics.

  • Discover how AI removes traditional loop bottlenecks by automating content creation, reducing activation friction, and enabling new forms of value creation.

• Lesson 1.2: AI-Powered Content Loops

  • Analyze LinkedIn’s collaborative articles strategy that uses AI to convert consumers into creators at 10x traditional rates.

  • Learn how to narrow the consumer-to-creator canyon by using AI to lower creation barriers while maintaining content quality and uniqueness.

  • Develop strategies for maintaining content uniqueness at scale when everyone has access to the same AI tools.

• Lesson 1.3: Building Multi-Loop Growth Models

  • Master the three-step growth sequence of finding initial arbitrage (Spark), converting it to sustainable loops (Transform), and layering multiple loops (Build).

  • Design primary loops that drive core growth and secondary loops that reinforce and amplify your primary mechanism.

  • Understand how different loops interact, when they reinforce versus cannibalize, and how to sequence loop development for maximum impact.

Answer Engine Optimization (AEO)

• Lesson 2.1: The New Discovery Paradigm

  • Understand the shift from search engines showing 10 blue links to AI providing direct answers, fundamentally changing how users discover products.

  • Analyze how ChatGPT, Claude, and Perplexity are training users to expect immediate, synthesized answers rather than exploring multiple sources.

  • Quantify the $8.5 billion impact of AEO on digital marketing and develop frameworks for measuring your share of AI-generated recommendations.

• Lesson 2.2: Building Your AEO Strategy

  • Develop comprehensive question research methodologies that identify what your target customers ask AI assistants throughout their journey.

  • Build authority signals that AI systems recognize, including structured data, consistent NAP information, and strategic citation networks.

  • Create multi-format content strategies that serve different AI consumption patterns, from quick answers to detailed explorations.

  • Implement attribution systems that track both direct referrals and brand impact when users see you mentioned but don’t click through.

  • Scale content creation using AI assistance while maintaining the quality signals that influence AI recommendations.

Signal-Based Go-to-Market

• Lesson 3.1: The Signal Revolution

  • Move beyond demographic and firmographic targeting to identify real-time behavioral signals that indicate purchase readiness.

  • Transform outbound sales from high-volume spray-and-pray to precision targeting based on observable customer actions.

  • Study real examples from Vanta (SOC 2 signals), Rippling (multi-location detection), and Clay (business model identification).

• Lesson 3.2: Building Signal Discovery Systems

  • Identify high-intent behavioral triggers unique to your value proposition that competitors haven’t recognized or can’t access.

  • Build technical infrastructure for capturing, processing, and activating signals at scale using AI and automation tools.

  • Develop validation frameworks and confidence scoring systems to ensure signal quality before routing to sales teams.

Strategic Growth Sequencing

• Lesson 4.1: The Three Growth Levers

  • Master optimization of existing loops to extract maximum value from current growth mechanisms before adding complexity.

  • Understand when and how to add new loops, balancing the opportunity cost against optimization potential.

  • Deploy linear activities strategically as activation energy for loops or to capture high-intent, low-volume opportunities.

  • Learn the strategic hierarchy and why pulling these levers in the wrong order wastes resources and misses growth windows.

• Lesson 4.2: The S-Curve Framework

  • Identify launch phase opportunities where early adoption provides outsized returns despite high uncertainty.

  • Navigate growth phase dynamics where speed of execution matters more than perfect optimization.

  • Maximize maturity phase value through systematic optimization when channels become efficient but competitive.

  • Recognize decline phase signals and transition strategies before channels become unprofitable.

Retention and Engagement

From Static Journeys to Dynamic Experiences

Use Case Evolution

• Lesson 1.1: Rethinking the Use Case Map

  • Discover how AI dramatically expands the problem space by making previously impossible or uneconomical use cases suddenly viable.

  • Learn how AI democratizes professional skills, enabling non-experts to access capabilities that previously required years of training.

  • Master the permission-to-frequency shift where AI transforms occasional use cases into daily habits by removing friction and expertise barriers.

• Lesson 1.2: Use Case Sequencing Strategy

  • Design strategies that guide users from simple initial use cases to high-value core use cases that drive long-term retention.

  • Manage the complexity of expanding AI capabilities that continuously unlock new use cases without overwhelming users.

  • Build discovery mechanisms that help users naturally find and adopt adjacent use cases as their comfort and sophistication grow.

Intelligent Activation

• Lesson 2.1: Activation Foundations

  • Understand activation as the critical bridge between your marketing investment (acquisition) and product investment (value delivery).

  • Identify and optimize setup moments (getting users ready) and aha moments (experiencing core value) in AI-enhanced products.

  • Master the four fits of activation (Audience, Promise, Urgency, Knowledge) and ensure alignment across all touchpoints.

• Lesson 2.2: AI-Enhanced Activation Patterns

  • Implement personalized onboarding at scale using AI to adapt experiences based on user characteristics and behavior patterns.

  • Deploy predictive setup assistance that anticipates user needs and proactively removes friction before users encounter obstacles.

  • Create dynamic value demonstrations that showcase the most relevant capabilities based on each user’s specific context and goals.

Proactive Engagement Systems

• Lesson 3.1: From Reactive to Predictive

  • Build behavioral pattern recognition systems that identify when users are ready for new features or at risk of churning.

  • Design anticipatory value creation strategies that solve user problems before they explicitly ask for solutions.

  • Develop context-aware intervention frameworks that provide the right help at the right moment without being intrusive.

• Lesson 3.2: AI Habit Formation

  • Apply environmental loop principles by embedding triggers where users already spend time rather than requiring new behaviors.

  • Leverage usage-based gamification mechanics that turn AI token limits into engagement drivers rather than friction points.

  • Optimize trigger strategies across channels, timing, and messaging to maximize habit formation without causing notification fatigue.

Intelligent Resurrection

• Lesson 4.1: Redefining Dormancy

  • Understand why traditional resurrection campaigns fail when they assume users left due to dissatisfaction rather than evolved needs.

  • Recognize capability-driven re-engagement opportunities where product improvements create new reasons for dormant users to return.

  • Develop frameworks for identifying which dormant users are worth targeting based on their original use case and current product capabilities.

• Lesson 4.2: Resurrection Through Evolution

  • Craft compelling narratives that communicate how your product has evolved since users last engaged without overwhelming them with features.

  • Time re-engagement campaigns based on capability milestones and market shifts rather than arbitrary time periods.

  • Define success metrics that account for users returning for entirely different use cases than their original engagement.

Monetization and Pricing Strategy

Capturing Value in Variable-Cost Economics

The AI Monetization Framework

• Lesson 1.1: The Four-Part Model

  • Master the shift from seat-based to usage-based pricing models as AI makes per-user pricing increasingly untenable.

  • Understand the spectrum of what you charge for, from features to capabilities to outcomes, and how AI enables outcome-based models.

  • Navigate dynamic pricing considerations where both your costs and customer value vary dramatically based on usage patterns.

  • Develop strategies for pricing timing in volatile markets where capabilities improve monthly and costs change weekly.

• Lesson 1.2: The Monetization Triad

  • Align customer value perception with pricing to ensure willingness to pay exceeds price while capturing fair value.

  • Integrate monetization with growth loops so pricing accelerates rather than inhibits viral, content, or paid acquisition mechanisms.

  • Ensure business model viability by balancing customer acquisition costs, lifetime value, and the volatile unit economics of AI products.

  • Find the sweet spot where all three triad components align to create sustainable, scalable monetization.

Value Metric Innovation

• Lesson 2.1: Beyond Traditional Metrics

  • Explore usage-based model variations including API calls, compute time, output volume, and custom metrics aligned with value creation.

  • Evaluate outcome-based pricing feasibility by understanding when you can reliably deliver and measure business outcomes.

  • Design hybrid models that combine usage, seats, and outcomes to balance predictability with value alignment.

• Lesson 2.2: Cost Structure Adaptation

  • Manage AI inference costs that scale with usage by understanding token economics and optimizing prompt engineering.

  • Implement intelligent caching, batching, and routing strategies to control costs while maintaining service quality.

  • Develop margin preservation strategies that account for improving model efficiency and declining API costs over time.

Packaging Strategies

• Lesson 3.1: Bundling AI Features

  • Recognize when AI becomes table stakes in your category and must be included in core offerings to remain competitive.

  • Navigate competitive parity considerations where matching competitor AI features is necessary but not sufficient for differentiation.

  • Design cost absorption models that bundle AI into existing tiers while preserving unit economics through usage limits or fair use policies.

• Lesson 3.2: Add-On and Standalone Models

  • Isolate premium AI capabilities as paid add-ons when they provide clear differential value to power users.

  • Target new buyer personas with standalone AI products that solve different problems than your core platform.

  • Leverage independent go-to-market advantages including focused messaging, specialized sales teams, and distinct positioning.Pricing Transitions

• Lesson 4.1: Research Methodology

  • Conduct comprehensive cost analysis that maps AI inference costs to user segments and usage patterns.

  • Execute willingness-to-pay research using Van Westendorp analysis adapted for AI value propositions.

  • Perform market anchoring analysis to understand competitive dynamics and customer reference points.

• Lesson 4.2: Execution Excellence

  • Study the Cursor pricing failure to understand how poor execution can turn your biggest advocates into vocal critics.

  • Design grandfathering strategies that balance fairness to early adopters with business model sustainability.

  • Develop communication frameworks that explain pricing changes in terms of value creation rather than cost increases.

  • Manage customer relationships through pricing transitions with transparency, empathy, and clear value demonstration.

Building AI-Era Defensibility

From Static Moats to Dynamic Advantages

Foundations of AI Defensibility

• Lesson 1.1: Misconceptions of Defensibility in AI

  • Understand why speed alone isn’t defensibility but rather a bridge to distribution opportunity, and how to sequence from speed to sustainable advantages.

  • Recognize that distribution is a stepping stone rather than a destination, requiring active conversion into deeper lock-in mechanisms.

  • Evaluate when data creates genuine marginal value versus redundant information, distinguishing between data accumulation and true data moats.

  • Learn why all moats are time-bound bridges rather than permanent fortresses, requiring continuous evolution to the next defensive position.

  • Discover how traditional moat types don’t disappear but evolve and adapt to new AI realities, maintaining relevance in transformed forms.

• Lesson 1.2: The 5 Core Types of Defensibility

  • Master the timeless categories of defensibility (direct network effects, cross-side network effects, data network effects, brand, economies of scale) and how each manifests differently in AI contexts.

  • Understand how data network effects have elevated from the weakest to potentially the strongest form of defensibility through AI’s ability to transform data into capabilities.

  • Analyze the progression from social graphs to algorithmic distribution to AI-generated content and what this means for network effect design.

  • Evaluate which variations of traditional defensibility patterns will thrive versus collapse as AI capabilities expand.

• Lesson 1.3: Defensibility Stacking

  • Apply the Motte-and-Bailey framework to build bailey defensibilities (speed, distribution hacks, early data advantages) while constructing motte defensibilities (true network effects, brand, scale).

  • Navigate the three-phase journey from Land Grab (achieving escape velocity) through Fortification (building first true moat) to Dominance (creating compound, reinforcing moats).

  • Study success patterns like LinkedIn’s progression from direct network effects to cross-side effects to data network effects versus failures like Groupon’s confusion of virality with defensibility.

  • Use the AI Defensibility Decision Matrix to assess your position across speed and moat depth, determining strategic priorities and survival probability.

  • Design bridge strategies where each form of defensibility enables the next, creating a deliberate sequence from speed to distribution to engagement to data to network effects to platform.

Data Network Effects in AI

• Lesson 2.1: The AI Data Network Effect

  • Build the flywheel of aggregating audience → generating data → improving AI experience → acquiring more audience, creating sustainable advantage over time.

  • Understand why proprietary data and functionality matter more than ever when AI capabilities are commoditizing and horizontal platforms threaten specialized tools.

  • Leverage AI to transform static data into dynamic insights through pattern recognition, personalization engines, and predictive capabilities.

  • Identify the limitations of horizontal AI platforms (lack of domain depth, real-time data, workflow integration) that create opportunities for specialized solutions.

  • Avoid common mistakes like overestimating data value, neglecting continuous loops, and missing opportunities to combine multiple assets.

• Lesson 2.2: How to Evaluate Your Data

  • Assess marginal value as the key metric, determining whether your data significantly changes model weights or outputs beyond standard training sets.

  • Evaluate exclusivity and accessibility factors including login walls, replication difficulty, and whether data is user-specific versus publicly available.

  • Measure freshness and timeliness considering decay rates, real-time access, and ability to maintain currency better than competitors.

  • Analyze depth and comprehensiveness in specific domains, specialized information coverage, and completeness relative to competition.

  • Build reinforcement capabilities through high-quality feedback collection, continuous improvement mechanisms, and translation of user behaviors into AI enhancements.

• Lesson 2.3: How to Build Your Data Network Effect

  • Execute the three-phase framework starting with strategic public data curation (like MidJourney’s quality-focused approach) to create immediate differentiation.

  • Implement refinement through RLHF and feedback loops that create truly proprietary data from user interactions competitors cannot access.

  • Pursue expansion through strategic partnerships and data aggregation once you’ve established critical mass and proven value.

  • Design user incentives for data contribution through immediate value exchange, progressive personalization, and community benefits.

  • Build implementation roadmaps that sequence foundation (curation) → refinement (feedback) → expansion (partnerships) with clear success metrics at each phase.

Network Effects and Brand Evolution

• Lesson 3.1: Direct Network Effects with AI

  • Understand how AI challenges the fundamental assumption that products need other humans to create value, potentially replacing human actions entirely.

  • Analyze the progression from friend graphs (Facebook) to algorithmic distribution (TikTok) to AI generation (Character.AI) and its implications.

  • Apply the AI Substitution Test to determine whether AI can perform your value-creating actions and whether it provides superior value through infinite availability and perfect personalization.

  • Identify which direct network effects survive (real-world activity documentation, authentic relationships, team coordination) versus collapse (content creation, information synthesis, basic validation).

  • Design strategies that either resist AI by doubling down on authentic human connection or embrace AI to build new types of defensibility.

• Lesson 3.2: Cross-Side Network Effects with AI

  • Master the three core marketplace functions (discovery, comparison, transaction facilitation) and how AI threatens each differently.

  • Navigate the three attack vectors of discovery agents (funnel collapse), transactional agents (disintermediation), and supply agents (direct integration).

  • Understand how AI compresses the traditional marketing funnel into single interactions, breaking LTV-based customer acquisition math.

  • Evaluate defensibility across discovery (exclusive supply, real-time availability), comparison (proprietary signals, domain expertise), and transaction facilitation (trust infrastructure, regulatory compliance).

  • Design marketplace strategies that focus on becoming trust infrastructure and operational complexity that AI cannot replicate through software alone.

• Lesson 3.3: Brand with AI

  • Recognize how brand has elevated from weak defensibility to critical differentiator as functional advantages become easier to replicate through AI.

  • Build trust through addressing AI-specific anxieties around hallucination, data privacy, and alignment uncertainty.

  • Accelerate category ownership and creation by moving fast to define new AI-enabled workflows before competitors establish positions.

  • Leverage the product as brand ambassador where AI personality becomes the primary touchpoint and every interaction reinforces brand values.

  • Cultivate community-driven brand evolution through creator empowerment, shareable outputs, and platforms that gamify and amplify user-generated success stories.

Enroll In AI Growth Course

Or view all our courses such as AI Foundations, AI Productivity, AI Strategy, AI Leadership.

AI Growth | AI Leadership | AI Strategy | AI Foundations | AI Productivity | All Courses

It’s been 10 years since I first wrote about the Four Fits. It’s the most popular piece I’ve ever written and something I refer back to in almost every strategy conversation I have.

While the framework still holds, AI has massively changed each of the four elements as well as how they fit together. It was clear that I’d need to update it based on AI and its halo effect, but I only just now feel that I have enough clarity to do so. In some ways, it’s amazing it lasted ten years without needing an update. Now, it’s obvious that it does and I hope this refresh helps you sort through some of the noise out there today.

In this post, I’m going to run through the Four Fits. I’ll summarize the framework, look at how some of the best companies of the AI era lean into the model, and talk through a few important ways that AI alters it.

The Four Fits: A quick refresher

Before we examine how AI disrupts the Four Fits, let’s establish our foundation. In order for a company to reach $100 million at venture speed, it must achieve all four. As I wrote in the original piece, when all four click, growth comes easy:

These companies grow despite having organizational chaos, not executing the “best” growth practices, and missing low-hanging fruit. I refer to these companies as Smooth Sailers - a little effort for lots of speed.

In other companies, growth feels much harder. It feels like pushing a boulder uphill. Despite executing the best growth practices, picking the low-hanging fruit, and having a great team, they struggle to grow. I refer to these companies as Tugboats - a lot of effort for little speed.

Everyone knows about Product Market Fit. And while it is one of the Four Fits, it’s only one of them. In the startup world, truisms like “Product Market Fit is the only thing that matters” have become common. Again, important but not the end-all-be-all.

Successful products require alignment across four distinct but interconnected dimensions. These are the Four Fits:

• Product Market Fit occurs when you build something that a meaningful segment of customers desperately wants. This is the classic definition most product leaders know. As Marc Andreessen wrote, you know you have it when “the customers are buying the product just as fast as you can make it -- or usage is growing just as fast as you can add more servers.”

• Product Channel Fit recognizes that products must be built for the channels where customers discover them (channels do not mold to products). TikTok wasn’t just built for short-form video—it was specifically architected for mobile consumption and social sharing. Pinterest, TripAdvisor, and others were architected specifically for search. The products reflect the channel’s constraints and opportunities.

• Channel Model Fit ensures your business model economics work with your chosen distribution channels. A freemium product with $10/month pricing can’t succeed through enterprise sales teams, just as a $50,000 annual software license can’t rely on viral social media growth.

• Model Market Fit confirms that your chosen business model aligns with how your target market prefers to buy and pay for solutions. Most importantly, that alignment adds up to a $100 million+ potential revenue product (e.g. 100,000 customers paying $1,000 per year).

You can read the original series here: The 4 Growth Frameworks You Need to Build a $100 Million Product.

How AI changes growth: product, market, channel, model

In the first post of that series, I wrote that, “The fits are always evolving/changing/breaking. When that happens, you can’t simply change one element, you have to revisit and potentially change them all.” Change was always anticipated but on a timeline that provided enough time to analyze and react. There are several changes influencing the Four Fits right now, and they are happening at unprecedented speed.

Here is an overview of each of the components and the major underlying change in each.

Product

AI increases both the problem space and the solution space. You can now solve problems that you couldn’t previously and you can offer new solutions to old problems that couldn’t previously be solved

And these aren’t incremental possibilities, they are exponential. Like I said earlier, LLM capabilities are doubling about every seven months, which means the problem and solution space will continue to increase dramatically. Just imagine a scenario where a task that currently takes your customers eight hours to complete is handled entirely by an agent. Then think about 40 hours, 80 hours, etc. This is the type of change that’s coming over the next few years.

As you expand to solve new problems, you’ll need to find new Product Market and Product Channel fits. It’s not a given that more or better solutions translate directly into a fit.

Market

AI is expanding some markets and shrinking others. Canva always pitched itself as a design tool for everyone, but its TAM is exploding. Its free plan says it’s “For designing or working on anything.” And now that users can describe a design rather than dragging shapes and picking colors themselves, it’s actually true. AI app building tools like Lovable and Bolt turned the niche “no-code” market into a massive one.

On the other side, the homework help tool Chegg watched its Product Market Fit collapse in just a few months. Seat-based customer support tools are suddenly selling to companies with fewer support employees. Today, AI augments support but for some companies, it could eliminate the human factor altogether.

My podcast co-host Fareed Mosavat says that so many of the things we used to count on feel like quicksand. Markets reflect customer expectations, and those are changing quickly. If a person uses and likes ChatGPT, it changes their expectations of other products too.

One other change in markets is that AI democratizes skills that were usually reserved for specialists. AI can help non-technical people create apps, non-designers generate beautiful images, or someone who has no video editing skills make an eye-catching video. This democratization of skills expands markets and potential TAM for many products.

Channel

As I wrote in The Next Great Distribution Shift, we’re living through one of the most significant technology shifts in history, yet we’re still distributing products like it’s 2015. SEO traffic is down across the board. Social sites are walled gardens. PPC platforms are cashing in on the bleakness. In other words, the old playbooks are burning.

This is partially due to AI but also part of a natural cycle where companies incentivize users with free/cheap distribution and slowly close it off over time. The cycle looks like this:

• Identify the Moat: Every platform starts by figuring out what will make it unassailable. The social graph for Facebook, search data for Google, and an app ecosystem for Apple.

• Open the Gates: Next, platforms create “open” ecosystems, practically begging developers to build on top of them. Free API access. Viral growth mechanics. Revenue sharing that seems too good to be true.

• Close for Monetization: Eventually, the rules change. What was free becomes paid. What was permitted becomes restricted. The platform starts competing with its own developers, often killing the very businesses that helped it grow.

We’re at a moment in time when most of the best distribution options are firmly in the third phase of this cycle. But a change is coming and I predict that ChatGPT will be the next big distribution opportunity, both through AEO/discovery and possibly a developer platform.

Model

Your model is (1) how you charge, (2) when you charge, (3) what you charge for, and (4) the amount you charge. There was plenty of nuance here prior to AI and most companies are still tinkering with pricing to figure out how to monetize AI. There are at least three new dynamics at play:

• Cost to serve is increasing: Many SaaS tools relied on PLG partially because the cost to serve free/freemium users was marginal. LLM costs vary wildly and can be significant. Bearing this new cost can break the economics of free/freemium products.

• Token usage is increasing: Heavier tasks eat up more tokens. Additionally, users overwhelmingly prefer the latest LLM models, even though previous ones are far cheaper. As TextQL founder Ethan Ding wrote, “GPT-3.5 is 10x cheaper than it was. It’s also as desirable as a flip phone at an iPhone launch.” Falling tech costs won’t fix your economics because they aren’t falling.

• Willingness to pay is up in the air: The magic of some AI tools is wearing off. It’s unclear exactly how users will perceive the value of products like this as the initial awe fades into a new baseline expectation. You can pretty quickly gauge the change in sentiment on the Cursor subreddit. It looks something like this:

Customer demand for AI is insanely high, but that doesn’t automatically translate into profit or sustainable businesses.

What happens when AI changes everything all at once?

I can’t emphasize enough how much each of these fits influences each other. I’ve just described them in isolation but in practice they all must be in harmony. Without that harmony, you have almost no chance at becoming a $100 million+ company.

Let’s take a look at how each fit is changing.

How Product Market Fit changes

Companies can find and lose product market fit almost instantly. It’s important to remember that Product Market Fit is only a moment in time. Changing market conditions and new technology have always forced companies to refine their offerings to keep the fit. But changes were gradual enough that they had time to assess and roll out features to keep pace.

For example, in the mobile shift:

• New technology capabilities happened in yearly cycles.

• It took time for an ecosystem of developers to emerge around mobile.

• It took years for a big enough audience of people to adopt the devices and have fast enough data plans for the new use cases to emerge.

The time between new tech capability and customer expectations increasing was multiple years at a minimum. This “slow” acceleration gave companies breathing room to adapt.

But something different is happening with AI. The time between these steps in AI is much faster.

• New AI tech capabilities are being launched monthly, if not weekly.

• There is a massive ecosystem of developers around this new tech already and growing.

• Distribution of this tech to hundreds of millions has happened fast, cheap, and freely.

As a result, we don’t see the Product Market Fit threshold accelerate, we can see it inflect. And when that happens, a company can lose Product Market Fit overnight.

Chegg illustrates this well. In January 2024, it was valued at $1.2 billion. By October 2024 (9 months later), it was valued at $150 million. A 90% decline in nine months, losing a half a million subscribers in that time. It went from essentially break even in 2023 to losing $600 million in Q3 2024 alone.

How Chegg’s product market fit collapsed

Chegg’s primary subscription offered homework help to students. The main value prop was high-quality answers written by curated humans. If we look at it through the lens of its growth model, its core growth loop is a company-generated, company-distributed content loop. More quality answers were then distributed via SEO and other channels, which led to more subscribers and engagement, which led to funding more quality answers.

But when OpenAI launched ChatGPT, students could just enter their homework and get an immediate personalized answer. The answer wasn’t always right, but the fact that it was instant and free created both a 10x value prop and immediate distribution.

It broke the growth loop. As subscribers started to churn, Chegg could fund less company-generated content, which leads to less new subscribers and engagement, which leads to less money, which leads to funding less company-generated content.

It’s not uncommon for companies to have to innovate their way out of changing dynamics, but it’s happening too fast right now. Chegg has responded on a pretty typical timeline. It launched its first AI features 26 months after ChatGPT launched. In the past, that may have been a fast enough response, but not in the AI era.

How Product Channel Fit changes

The second of the four fits is Product Channel Fit. All successful products are built off the back of another channel. TikTok leveraged Facebook Ads, Booking.com leveraged Google Ads, and HubSpot leveraged Google SEO. Product Channel Fit states that products are built to mold to channels. Channels do not adapt to your product. That is because you do not have control of the channels where your audience lives. You play by their rules.

There is an entire ecosystem of $1 billion+ companies built on the back of long-tail SEO. Pinterest, TripAdvisor, Quora, G2 and Quora just to name a few. At the center of their growth model is a core growth loop: User Generated, Company Distributed Content Loop:

• Step 1: A new user signs up (or existing user returns)

• Step 2: They create a piece of content on the platform (a review, a pin, etc.)

• Step 3: Company optimizes pages of that content to be indexed by Google.

• Step 4: A new or existing user finds that piece of content in a search (repeat step 1)

This isn’t just B2C companies either. B2B companies like G2 and HubSpot have built the majority of their growth engine around similar loops. All these companies have spent 10+ years tuning and optimizing their products to tap into SEO.

What happens if the channel loses product market fit?

Here’s the risk companies should be aware of today: Imagine a scenario where a major platform like Google suddenly loses its Product Market fit. The ripple effects would be dramatic and far-reaching. Companies like TripAdvisor and Pinterest, which rely heavily on search traffic, would face immediate existential challenges. This disruption would then cascade down to thousands of smaller businesses that depend on these intermediary platforms.

At the time of this writing, we’ve seen some declines in some of these channels, Google especially. We have not seen the instant collapse. However, we’ve already seen that AI can have a rapid, nearly instant impact when it “clicks.” So, companies need to pay close attention to their distribution channels and prepare for disruptions.

What if a new channel emerges?

There’s a second scenario that companies often miss. Your existing channel doesn’t need to tank for you to face disruption. What if a major new channel emerges and your users start forming entirely new habits there?

When Facebook’s platform launched in 2007, people didn’t stop visiting web game portals like Webclip and Yahoo Games. But there was suddenly a new place where people were spending significant time and building habits. Web game companies tried to copy and paste their games into Facebook’s environment and failed completely. The games weren’t built for how people behaved in that social context (playing with friends rather than solo).

The same pattern repeated with mobile. Social game companies attempted to port their web experiences directly to mobile devices. It didn’t work because mobile usage patterns were fundamentally different.

The ChatGPT disruption is happening right now

As I described in The Next Great Distribution Shift, we’re seeing this exact pattern unfold today with AI. For many product categories, people are starting their research and discovery process on ChatGPT instead of traditional search engines or category-specific sites. This represents a fundamental shift in user behavior and starting points.

The implications vary dramatically by industry. Companies selling high-consideration purchases like travel or cars face immediate pressure. Users are conducting extensive research on ChatGPT before ever visiting traditional booking sites or dealer websites. These companies must figure out how to extend their products into AI channels or risk losing customers to competitors who do.

You’re vulnerable even when your channels are stable

You can be disrupted even if your existing channels continue performing well. If your audience shifts significant time and attention to new channels, you need to figure that channel out ASAP. Otherwise, you leave yourself open to being disrupted by a new player that figures out that channel first.

How Channel Model Fit changes

The third fit is Channel Model Fit. In Channel Model Fit, the channels that are viable for you to use are determined by your monetization model (how you charge, when you charge, what you charge for, how much you charge).

We’ve historically explained this through the ARPU ↔ CAC Spectrum. Every business lives on the ARPU ↔ CAC Spectrum. On one side, you have businesses with low ARPU and, as a result, have to use low CAC channels to drive customers. On the other side, you have businesses that have high ARPU and are therefore able to use high CAC channels.

There are a couple of ways that Channel Model Fit could collapse:

• Cost increases break GTM: One of the new problems that AI has introduced in monetization is meaningful costs with every use of the product. For example, freemium products will have difficulty introducing meaningful AI features as part of the free tier without drastically increasing the costs to support those free users. Traditional SaaS gross margins were typically above 70%. The Information reported that Lovable’s gross margin is 35% but it doesn’t include the cost of serving free users as part of that, which would drive it much lower. Depending on the product, this can break Channel Model Fit.

• Channel collapses could trigger monetization changes: Many products heavily rely on SEO traffic to drive inexpensive traffic to their low-priced or free products. But as the amount of organic traffic decreases, it may force companies to shift to other more expensive channels like paid advertising. They will then need to change their monetization model to align with the increased cost of those channels. This could be especially challenging when paired with increased costs to serve.

• Shifts in user adoption behavior: AI-powered innovations can create instant shifts in user behavior. For instance, if an AI plugin or chatbot reduces the friction of discovering and trying new solutions, potential customers might expect a near-zero friction trial. But higher-priced or more complex products still demand lengthy demos or a big sign-up fee. Almost overnight, your tried-and-true channels (e.g., lengthy content marketing funnels) might become far less effective. AI can fundamentally change where and how users make purchase decisions, invalidating established channel strategies.

Channel Model Fit hinges on having pricing and ARPU that align with the channels you use to acquire customers. If AI instantly upends that balance, you could end up:

• Stranded in the “Danger Zone.” If your ARPU is too low to afford human-powered sales, but too high/complex to succeed on low-cost viral or paid channels, you end up stuck.

• Massive churn or halted growth. If a competitor’s AI-driven product is cheaper or drastically more effective, your customers may churn en masse. Meanwhile, your inability to adapt channels quickly can stall new customer acquisition.

• Ballooning CAC and negative unit economics. In an effort to respond, you might try layering on AI consultancies or bigger marketing campaigns. But if you can’t recoup that expense within a year, you burn through capital fast. This negative cycle jeopardizes your company’s viability.

• Forced pricing and product overhauls. In a scramble to regain fit, companies may attempt to re-price overnight or re-bundle their offerings. But these rushed changes often don’t seamlessly align with the channels you’ve built—leading to confusion in the market, internal misalignment, or suboptimal “patchwork” channel strategies.

• Increased competitive pressure to move up or down market. If the AI wave commoditizes your existing segment, you might have to move up-market and serve enterprise customers, but that requires building out (and affording) a high-touch sales team. Or, you might need to go down-market—lower ARPU and rely on low-friction, product-led channels. Either direction is a fundamental pivot that can be risky, costly, and time-consuming.

The companies that survive this transition will be those that monitor their Channel Model Fit closely and can pivot quickly when the fundamentals shift. The key is building flexibility into your monetization strategy before you’re forced to change it under pressure.

How Model Market Fit changes

Model Market Fit is the concept that your market (and number of customers within your market) influence your model. Years ago, Christoph Janz from Point Nine Capital wrote a great post called The Five Ways To Build A $100 Million Business. It describes the number of customers you need at different price points to grow to $100 million.

Most companies end up falling into one of five areas that Christoph named:

• Elephants: Products that get 1,000 customers paying $100k+ a year. These are typically products built for enterprise customers like ServiceNow.

• Moose: Products that get 10,000 customers paying you $10k+ per year. These are typically products built for the mid-market like HubSpot.

• Rabbits: Products that get 100,000 customers paying $1k per year. These are typically products targeting small businesses like SurveyMonkey, Mailchimp, or Gusto. Products targeting consumers at high value moments also live here. For example, companies in real estate, insurance, etc.

• Mice: Products that get one million customers paying $100 per year. These are typically products that target prosumers like Dropbox, or companies that are subscription e-commerce like Ipsy or Dollar Shave Club.

• Flies: Products that get 10 million customers generating $10 per year typically via ads. Facebook, Snapchat, Buzzfeed, etc. all live here.

This can be visualized in this chart:

The diagonal line represents $100m in revenue. You can pinpoint your own business on this chart by finding your ARPU on the Y-axis and the number of customers you have on the X-axis. If you’re above the line, you have Model Market Fit. If you’re below it, you don’t.

AI breaks the math

AI is expanding some markets and shrinking others. This changes the math. Customer support teams have drastically shrunk because AI has replaced a lot of the work. Vendors selling seats suddenly have way fewer people to sell to, so the math of customers times ARPU breaks down.

But this creates an opportunity for a tool like Intercom’s Fin, an AI support bot, that monetizes based on resolved tickets. At 99 cents each, it needs to resolve just over 100m tickets per year to find Model Market Fit. (As a side note, as of August 2025, Fin resolves more than 1 million tickets per week.) If Fin sold seats to customer support teams, it’d likely collapse but AI has enabled a new solution that didn’t previously exist.

Are bigger markets always better?

Some markets are absolutely exploding. Companies like Cursor and Windsurf are tapping into a much larger TAM. AI writes the code, meaning a lot more people can develop products. Lovable, Bolt, Replit and v0 take this to another level and are tapping into a new market altogether. Building apps without any code at all creates an almost infinite TAM.

Replit grew revenue 50x in the last 12 months and has 40 million users and 175,000 paying customers, according to Sacra. It found escape velocity with a product-led growth motion and massive interest in vibe coding. Since then, it’s turned its attention towards team and enterprise, looking to gain traction as a B2B utility for product and design teams.

This is a case where a big market is the spark, but only 0.44% of Replit users actually want to pay. So while freemium plans can drive tons of free usage, (1) that usage can be expensive and (2) it is not necessarily a clear indicator that users will be compelled to pay for more.

Carta reports that 1,700+ startups shut down in 2023 and 2024, a significant uptick over previous years. Good ideas, great products and even plenty of interest from users ≠ long-term success.

The buyer persona shift

AI creates new buyer personas with different authority, timelines, and success metrics. The traditional IT buyer focused on features and integrations. The AI buyer often comes from operations, evaluating against headcount savings and process automation. Their approval processes, budget sources, and implementation timelines operate on completely different cycles than traditional software purchases.

Many of us have gotten so used to jumping around from one free/freemium AI tool to another that it’s hard to imagine a procurement team choosing a single tool to roll out to an entire company. But that’s already happening and it’ll shift buying power from early adopter types to CTOs.

3 ways AI causes Model Market Fit collapse

AI has the power to unravel one or more variables in the Model Market Fit equation in three ways:

• Dramatic reduction in ARPU: AI-driven tools often automate and commoditize what was once a complex or high-value service. For instance, if you’ve been charging $1,000/year for a product (like specialized analytics or content generation), a new AI-based competitor might drop the average price to $100/year or even free (supported by ads or data-collection).

• TAM contraction or fragmentation: Sometimes AI doesn’t just reduce prices, it also fundamentally changes who is a viable customer. Suppose your product’s main market was companies needing data entry. If AI automates large chunks of data processing, then suddenly the market itself (i.e., the set of customers who need your old solution) shrinks drastically.

• Serviceable addressable market (SAM) contraction: Even if the market is still there and your ARPU remains the same, AI can create new winners with network effects or brand-new approaches, making it harder for incumbents to keep (let alone grow) market share.

Rabbits are becoming Moose in some cases, while Mice turn into Flies in others. The change to become a $100 million business still exists, but the path may look different.

Risk on one side, opportunity on the other

This is all happening so fast. I hope this post triggers at least some urgency on your part. For all the risk out there, windows of massive opportunity are also opening. The move right now is to analyze Product, Market, Channel and Model for your business, then look closely at how each of the Four Fits is working, or not.

The Four Fits are a significant part of our new AI Growth course. It’s a massive overhaul of our 10-year-old Growth course, which has been taken by tens of thousands of product builders and has been a staple of our business for a decade now. You can enroll here.

Reforge’s Fall Cohort of AI courses starts October 14th, including guests from OpenAI, Canva, GitLab, Descript, Laurel and many more. Enroll now to secure a spot.

AI Growth | AI Leadership | AI Strategy | AI Foundations | AI Productivity | All Courses

Atlassian recently made headlines when it acquired The Browser Company for $610 million in cash. The Browser Company makes Arc and Dia, with the latter being the focus of this deal.

The M&A market is both active and frothy right now as companies jostle for positioning in a quickly changing market. Atlassian has a rich history of M&A, but this one hit different.

The world was different back in 2023 when Atlassian acquired Loom for $975 million. We broke that acquisition down on an episode of Unsolicited Feedback. That was largely viewed as a pragmatic move. Loom fit nicely into Atlassian’s suite of asynchronous collaboration tools and created a new entry point for users. The same goes for the 2017 acquisition of Trello. These tools gained escape velocity thanks to the consumer-grade experience, strong growth loops and freemium price point.

But even back in 2023, Atlassian’s flagship products Jira and Confluence were starting to gather some “enterprise dust.” Each was and is good software and a profitable line of business, but as the company moved upmarket, it couldn’t rely exclusively on product-led growth (PLG) anymore. In 2020, enterprise sales accounted for 15% of sales. By Q1 2025, it’s >40%.

As so often happens, this changed the sentiment around the products too. Over the last few years, Jira and Confluence’s relevance has faded among small teams and startups who gravitated en masse towards Linear and Notion.

You can see the Innovator’s Dilemma from a mile away. The company that was a pioneer of the product-led growth model was starting to lose a sliver of market share to upstart competitors. But if a new generation of startups prefer Linear and Notion, what happens when those companies grow up?

It’s relatively easy to work backwards to the strategic reasoning behind the Loom acquisition. Atlassian wanted a product with viral growth potential that also passes the “toothbrush test,” meaning there’s a reason to use it every day. And the deal has proven to be a good one. In its August 2025 earnings, Atlassian reported that Loom’s monthly active users are growing more than 30% year-over-year.

Here are five smart reasons why Atlassian may have acquired The Browser Company

The Browser Company deal isn’t as straightforward. Browsers are among a small handful of the most important tools people use on a daily basis, but have mostly been dominated by the tech’s biggest players, namely Google and Apple. Where Loom fit nicely into Atlassian’s existing suite, The Browser Company acquisition is a much bolder bet on a future that looks very different than today.

Tech Twitter had a field day with this news. It was highly entertaining, but mostly poked fun at the deal. Atlassian is a legendary SaaS company, so I analyzed this deal under the assumption that the people making it are smart, thoughtful, and deliberate. Knowing plenty of people who work or have worked at Atlassian (including many Reforge alums), I know this to be true.

Here are the five reasons this deal may make sense.

1: Agents are an existential threat to Atlassian’s core business model

Atlassian built its strategy around a simple playbook. JIRA becomes the starting point for development teams, then it can cross-sell Confluence, Bitbucket, Trello, and other tools. This worked because JIRA owned the daily workflow for millions of developers and project managers.

That foundation is shifting. New AI-powered starting points are building daily habits with users. ChatGPT has become the first stop for many knowledge workers. Claude handles complex analysis. Cursor is becoming the go-to environment for AI-assisted coding.

These aren't just new features—they're becoming the new entry points for how people work. When you start your day asking ChatGPT to help plan your sprint or use Cursor to write code, you're bypassing traditional SaaS applications entirely.

The math is simple for Atlassian. If agents become the primary interface for work and those agents are controlled by OpenAI, Anthropic, or other AI companies, then owning JIRA matters less. You can't cross-sell adjacent products if users never visit your core product.

ChatGPT, Claude, and Cursor have already claimed significant mindshare as AI starting points. Rather than compete directly in that space with little chance of winning, it placed a bet on browsers as an alternative entry point that might still be winnable and have room for innovation.

2: B2B browsers create a new aggregation layer for enterprise work

The hypothesis is that AI will create separate B2B and B2C browser markets, and this positioning could be extremely valuable.

First, enterprise data and context requirements are fundamentally different. AI applications need massive amounts of data and context to produce valuable outputs. Do companies want that sensitive business data stored in browsers employees also use for personal browsing? Enterprises need control over where this information lives and how it flows between applications.

Second, user behavior has already shifted. Most people don’t interact with personal applications through desktop browsers anymore because that’s moved to mobile. Your desktop browser usage is primarily work-focused now, but Chrome and Safari were built for general consumer web browsing, not enterprise workflows.

But there's a bigger opportunity here. The browser sits at a unique position in the technology stack: one layer above individual applications like ChatGPT or Slack, but one layer below your operating system. This gives it potential to coordinate across everything you do at work.

Think about how Apple leveraged this positioning on mobile. By controlling the device layer, it can coordinate experiences across apps in ways individual app makers cannot. A B2B browser could play a similar coordinating role for knowledge work.

Imagine a new employee joining your company. Instead of logging into fifteen different applications and manually providing context to various AI agents, they spin up a pre-configured browser with access to approved applications, relevant data, and connected identity systems.

This represents the shift from "systems of record" to "systems of action." The places where you take action become more valuable than where you store information. Atlassian already serves enterprise customers, understands compliance needs, and has established IT relationships, which are advantages that consumer browser makers can't easily replicate.

3: Atlassian needs bigger bets to keep growing

Atlassian has reached the scale where incremental moves won't generate the growth it needs. Its main strategy for years has been adding adjacencies to the JIRA suite—Confluence for documentation, Trello for project management, Loom for video communication, etc.

This adjacency playbook works up until a point, similar to how HubSpot added marketing automation, CRM, and service tools around its core marketing product. But there's a limit to how many adjacent tools you can sell to the same customer base.

At its current scale, it needs whole new S-curves for growth. Companies exist to grow, and if they're not growing, they immediately shift into harvest mode. Atlassian may be entering its moonshot phase, similar to Google's "Other Bets" division.

The browser market represents massive potential upside. If you own one of the primary starting points for how people work, it’s an enormous business opportunity. Even if it’s a low probability bet, the potential return justifies the investment for a company with $3 billion in cash and strong free cash flow.

This isn't unusual for companies at Atlassian's stage. It’s moved beyond optimizing its existing products and into territory where it needs to find entirely new areas of business.

4: The "Atlassian Problem" - losing the next generation

We follow this exact pattern at Reforge. Our stack includes Linear for project management, Notion for documentation, and Slack for communication. The one Atlassian product we use is Loom, which it acquired specifically to address this problem.

This is a common lifecycle for SaaS companies. You start by serving early adopters and small companies without switching costs. Success drives you upmarket where the money is, but that opens the door for new competitors to capture the bottom of the market.

If new companies prefer different tools, what happens when those companies grow up? Atlassian risks becoming "rich but irrelevant to anyone building something new" but its leadership is aware enough to fight it.

The Browser Company acquisition follows the same logic as Loom. Get an entry point with next-generation companies in an adjacent space where you can potentially bridge back to your core products over time.

5: It’s not as expensive as it seems

The numbers put this deal in perspective. Atlassian has $2.9 billion in cash, so $610 million represents about 20% of its current cash position. It also generates about $1.4 billion in free cash flow per year. For a company with strong free cash flow and profitable operations, this is manageable risk for potentially massive upside.

$610 million sounds crazy on the surface because it's a lot of money, but it's actually a relatively small bet for a company of Atlassian's scale and financial strength. It’s always been contrarian: profitable at IPO, based in Australia, very low-priced, and had no sales team for years.

If browsers become a new starting point for work and Atlassian captures even a fraction of that market, the return would be enormous. Even if it's a low probability outcome, the potential upside justifies the investment. This is exactly the type of calculated risk a company at its stage should be taking.

Why the acqui-hire theory doesn't hold up

I've seen some people suggesting this was an expensive talent acquisition deal, similar to what we've seen with other AI-focused acquisitions. The facts don't support this theory.

First, the deal structure is wrong. Talent acquisitions typically use stock, not cash, to retain key employees through equity. Atlassian paid $610 million in cash, which suggests it’s serious about the product strategy rather than just acquiring a team.

Second, its public statements contradict this theory. Atlassian and The Browser Company founders have been explicit about their intention to "reinvent the browser for knowledge work." It’s hiring 50 more people and making clear investments in the product direction.

Finally, Atlassian doesn't have the talent attraction problems that would justify this approach. Unlike older enterprise companies that struggle to recruit top engineering talent, Atlassian remains an attractive employer that can compete for talent through traditional hiring rather than expensive acquisitions.

The talent at The Browser Company is impressive, but not irreplaceable at this price point. This deal is about product strategy and market positioning, not just acquiring a team.

It’s everyone vs. everyone

As we discussed on Unsolicited Feedback, it’s Everyone vs Everyone right now. The M&A market reflects how seriously companies are taking the AI transition. The frothy valuations and bold acquisitions we're seeing aren't just about current capabilities. These are about positioning for a fundamentally different future.

The mobile and cloud shifts created space for new players to build without stepping on each other's toes. AI collapses those boundaries. Companies that never competed before are suddenly fighting for the same customers as AI expands what each business can offer.

This is exactly what’s happening with Atlassian. I doubt it ever considered Google a direct competitor in the past, but it’ll go head to head in the browser wars.

A couple of weeks ago Anthropic launched a new Claude feature called “AI-powered apps.” There are some unique elements to this feature that creates a potentially new powerful growth model. and I broke it down on a recent episode of Unsolicited Feedback.

Listen On: Spotify or Apple Podcasts

What Are Claude AI-Powered Apps?

As Claude describes in the announcement, AI-powered apps in Claude are “the ability to build, host, and share interactive AI-powered apps directly in the Claude app.” It’s basically vibe-coding directly within Claude rather than using a purpose build tool like Cursor, Bolt, Lovable, Replit.

Examples

Here are a few examples:

• Educational Tools: A high school teacher created an SAT prep app that generates personalized practice questions based on each student's weak areas. Students log in with their Claude account, and the app adapts its difficulty in real-time.

• Business Utilities: A product manager built a requirements analyzer that reads uploaded PRDs and generates test cases, edge scenarios, and potential risks. Teams across her company now use it daily.

• Data Analysis: A sales ops team created an app that ingests their CSV exports and lets anyone ask questions in plain English: "Which territories are underperforming this quarter?" or "Show me deal velocity by product line."

Fareed in the Unsolicited Feedback episode highlighted something he built:

"I built an app for one of my kids to do SAT studying. Kind of a pain in the butt in something like Lovable, right? Because I got to get an OpenAI API key... I need to create accounts and all this sort of stuff. I'm not planning on charging for it.”

The platform's versatility and constraints lends itself to diverse range of functional and creative artifacts being shared by the community. What makes these apps special isn't that they were built without traditional coding (you can do that in a ton of places) – it's how they handle the details around authentication and sharing.

The Unique Economics and Constraints

Here's where Claude breaks from other AI app development tools. When someone uses your AI-powered app:

• They authenticate with their existing Claude account – No separate login system for you to build or maintain.

• Their API usage counts against their subscription, not yours – No worrying about managing model costs. If your app goes viral and gets a million users, you don't get a million-dollar API bill. Each user's prompts count against their own Claude subscription (Free, Pro, or Max). You pay nothing for their usage.

• No one needs to manage API keys – Forget about your users needing to enter their API key, usage limits, or security concerns. Claude handles all of that.

As you will see in this post, these constraints create a unique growth model for Anthropic and Claude.

The Core Growth Loop Of AI Powered Apps

Quick Growth Loop Recap

If you've been building products for any length of time, you know that sustainable growth rarely comes from one-off campaigns or viral moments. The products that win create systems that feed themselves – what we coined in 2017 as Reforge Growth Loops.

Think of growth loops as the compound interest of product development. Each cycle through the loop doesn't just add users; it multiplies your growth potential. While traditional funnels have a beginning and end, loops create a continuous cycle where outputs become inputs. You can see a lot of examples and materials here.

Why Growth Loops Matter:

• They create predictable, scalable growth without linear increase in costs

• They build defensibility through network effects and user-generated content

• They align user behavior with business growth naturally

The Three Core Types of Growth Loops

Understanding which type of loop you're building fundamentally changes how you execute:

1. Viral Loops – Users directly invite other users (think Dropbox's referral program or Zoom meeting invites)

2. Content Loops – Users create content that attracts new users (Pinterest pins, YouTube videos, Twitter threads)

3. Paid Loops – Revenue from users funds acquisition of new users through sales loops, paid marketing, and partnerships.

Each type has different constraints, metrics, and optimization strategies. Claude's AI-powered apps create something unique: an ai-powered content loop that creates potentially millions of other mini growth engines.

Claude’s Core Growth Loop For AI-Powered Apps

The core growth loop for Claude’s AI-powered apps looks similar to a User Generated, User Distributed Content Loop used by other companies like Eventbrite - A new user signs up, they create a piece of content, the user distributes that piece of content, which attracts new users.

In the case of Claude it is;

1. Step One: New Claude user signs up or returns to Claude

   A user signs up to Claude or comes back to use Claude.

2. Step Two: User creates an AI-powered app (no code required)

   That same user realizes they can solve a specific problem by building a lightweight application directly within Claude. Maybe it's a custom SAT tutor for their kid, a project tracker for their team, or a specialized calculator for their industry. The key? No technical skills required—just natural language instructions to Claude.

3. Step Three: Creator shares app with their network

   The creator distributes their mini-app to the people who need it. This could be internal sharing within a company, posting in professional communities, or simply sending a link to friends and family. The distribution is organic and targeted—people share solutions with those who have the same problems.

4. Repeat of Step One: Recipients must authenticate with Claude to use driving new or returning Claude users

   This is the crucial mechanism. Anyone who wants to use the shared app must sign up for Claude and authenticate through their system. The app won't work without it.

5. Repeat of Step Two: Some percentage of new users become creators themselves

   Once these new users are in the Claude ecosystem, they discover they can build their own AI powered Apps. The cycle keeps repeating

The incentives and constraints that Claude set up make this loop particularly interesting.

• Using Claude Auth - Using Claude authentication both makes it easier for the creator to create an AI powered application, but also drives users back into Claude’s ecosystem.

• End User’s Claude Subscription - Since the usage counts against the end user’s subscription w/ Claude, the creator doesn’t have to manage AI costs and complications. More importantly, it drives more use of the subscription probably driving upgrades for Claude.

The Compounding Universe Of Claude Apps

But here is where I think the growth model becomes particularly interesting. Each app built with Claude has it’s own set of growth loops driving users and usage to the application (and therefore Claude).

A project management tool built on Claude might have sharing and collaboration features that drive additional invitations. A specialized calculator might get passed around industry forums. Each app becomes a distribution channel for Claude itself.

This creates what we might call "nested loops"—the core creation-sharing loop at the platform level, plus individual sharing and engagement loops within each created application. The growth compounds not just through individual creators, but through the viral mechanisms built into their individual apps.

The Unique Constraints That Enable This

Creation Constraint

Every growth loop has constraints, bottlenecks that determine whether the system compounds or stalls. Most content loops fail at the creation step - converting enough users to actual creators. Getting users to produce something shareable requires overcoming significant friction: technical barriers, cost concerns, time investment, and unclear value exchange.

In content loops you see two types that work:

• Higher % conversion from user → creator with low distribution per piece of content

• Lower % conversion from user → creator with high distribution per piece of content

The math has to work between those two numbers in order to get the loop to compound.

Claude has designed some constraints in the system to encourage this creation for different group of the market.

The Creator Value Exchange

The magic lies in what creators get versus what they're willing to give up.

What creators get:

• Free hosting and infrastructure: Your app lives on Claude's servers with zero setup. No AWS accounts, no deployment pipelines, no server management. As Fareed Mosavat explains: "I built an app for one of my kids to do SAT studying. Kind of a pain in the butt in something like Lovable, right? Because I got to get an OpenAI API key... I need to create accounts and all this sort of stuff."

• No API costs or key management: When someone uses your app, their Claude subscription covers the API costs. You're not monitoring usage limits, managing billing, or worrying about runaway costs if your app goes viral.

• Built-in authentication system: User management becomes Claude's problem, not yours. No password reset flows, no security audits, no GDPR compliance headaches. Users authenticate once through Claude, and you're done.

• Future distribution potential through gallery: While not the primary value proposition, Claude is signaling a potential future discovery mechanisms through the Artifact gallery.

What creators give up:

• User ownership/data: You don't get full control over email lists, usage analytics, or the ability to communicate directly with your users outside of Claude's ecosystem. The relationship belongs to Claude (though they could add things to change this).

• Platform independence: Your app is tied to Claude's infrastructure. You can't easily migrate to another platform or offer it as a standalone service.

• Monetization options (currently): You still have to code in a way for people to pay/monetize though Claude could launch a payment system in the future. The value exchange is utility, not necessarily business building.

These constraints seem to target a different audience of potential app builders. On one end of the spectrum you have builders that want full control who use things like Cursor, Claude Code, etc. In the middle of the spectrum you have products like Bolt, Lovable, etc for less technical folks that are willing to give up some control. Claude creates even more constraints, enabling an even longer tail of application building for less technical people.

Lessons and Takeaways For Product and Growth Leaders

Let me be clear,I don’t know if this will actually work for a few reasons.

1. AI app development is crowded with purpose built products for the exact use case.

2. AI app creation within Claude will need to compete for attention of all the other things someone can do with Claude.

3. Claude's artifacts approach feels like the opposite end of the strategic spectrum from their other initiatives focusing on hardcore developers and enterprise customers.

But the attempt reveals several insights worth studying, regardless of outcome.

Growth Model Innovation

As Fareed mentions:

“They're basically turning code into content that can be distributed, shared, and drive new users, which I think is unique and new and very cool.”

AI is enabling new variations on growth loops

We're seeing growth mechanics that weren't possible before generative AI. The ability to create functional applications through natural language conversation changes who can be a creator and what creation looks like. Claude isn't the only company experimenting here. New technology capabilities often enable new growth mechanics (and new distribution platforms). When a fundamental capability shifts (like non-technical creation), the growth opportunities shift too.

Sometimes the constraint is the opportunity

Most platforms view constraints as problems to solve. Claude flipped this thinking. Instead of trying to compete with full-featured development platforms, they embraced severe limitations—no persistent storage, no external API calls, no direct monetization and used those constraints to enable a different value exchange.

The key is always how you align incentives across all participants in the loop

Claude's model works (if it works) because every participant gets immediate value:

• Creators solve their own problems first

• End users get utility without setup friction

• Claude captures user growth and engagement

This isn't revolutionary thinking, but it's easy to get wrong. Most platform strategies optimize for one participant at the expense of others.

Start constrained, expand deliberately

makes this point explicitly:

"I like what they've done with making this very, very constrained... I like that it's pretty small and narrow. I like that it really encourages tinkering and ease of use."

The temptation with new growth mechanics is to immediately build for every use case. Claude started with toy-like applications that encourage experimentation rather than business-critical development. This creates room to learn before the stakes get higher.

and I tackled this in our recent episode of Unsolicited Feedback. The TL;DR: If we're right, you have months (not years) to get your platform strategy right. The gates are opening. And the innovation replication cycle has never been faster.

In this post (and episode), we take some steps to help you learn how to play the game before the game plays you.

Listen On: Apple | Spotify | YouTube

Recap: The Next Great Distribution Shift

In The Next Great Distribution Shift, I laid out four points (skip this section if you’ve already read that article).

The AI Tech Shift Happened. The AI distribution Shift Is Just Beginning.

The AI technology shift has transformed product capabilities, business models, and competitive landscapes. Yet despite this technological revolution, we have not experienced its corresponding distribution shift. This isn't unusual. Every major platform transition follows the same pattern—technology first, distribution second.

Distribution shifts consistently lag technology shifts a couple of years. We witnessed this delay during the emergence of social networks, search engines, and mobile platforms. The technology foundation establishes first, creating new possibilities and disrupting existing workflows. The distribution revolution follows, determining which companies capture and control the value created by these technological capabilities.

We Are Now Approaching The Inflection Point Where The AI Distribution Shift Will Emerge

The competitive conditions necessary for distribution shift emergence are now aligned. We have achieved market consensus that AI chat experiences represent a massive, transformative category while the ultimate winner remains unclear.

Multiple major players are actively competing for platform dominance: OpenAI with ChatGPT, Anthropic with Claude, Google with Gemini, Meta with Llama, and Apple's strategic positioning remains uncertain. This competitive dynamic creates the exact environment that historically triggers distribution platform development.

Simultaneously, traditional distribution channels are experiencing significant degradation. SEO effectiveness has declined, app store discovery has become increasingly difficult, and paid advertising channels are delivering diminishing returns. This distribution scarcity creates market pressure for alternative channels, accelerating adoption of emerging platforms.

The convergence of competitive uncertainty and distribution scarcity establishes ideal conditions for platform emergence and ecosystem development.

The Distribution Shift Will Follow The Same 3 Step Cycle

Every successful distribution platform follows an identical progression that reflects structural market dynamics rather than platform-specific strategies.

Step 1: Moat Identification Market consensus emerges around category importance while multiple players compete for dominance. The eventual winner identifies sustainable competitive advantages that differentiate from feature parity competition. In the AI landscape, this moat is shifting from model intelligence to context and memory accumulation—the platforms that can gather and leverage user context most effectively will achieve escape velocity.

Step 2: Opening The Gates The leading platform creates an open ecosystem to accelerate moat development. This involves establishing value exchange mechanisms where third-party developers receive capabilities and organic distribution in exchange for extending platform functionality and contributing to competitive advantage accumulation. We are beginning to see early signals of this phase with integration announcements and platform development hiring.

Step 3: Closing The Gates Once competitive position is secured, platforms optimize for monetization and control. Organic distribution becomes artificially constrained toward paid channels, successful third-party applications get absorbed into first-party features, and revenue sharing terms deteriorate significantly. This phase is inevitable once platforms achieve market dominance.

There Is No Opting Out

Platform participation represents a strategic prisoner's dilemma where individual rational decisions create collective competitive dynamics that force market-wide participation.

If competitors integrate with emerging platforms and gain competitive advantages through enhanced capabilities or distribution access, non-participating companies face systematic disadvantage. Market forces compel integration regardless of long-term platform control concerns.

This dynamic explains why established companies like HubSpot integrate with ChatGPT despite obvious risks of user relationship displacement. The competitive pressure of potential customer preference for integrated experiences outweighs platform dependency concerns.

The strategic choice is not whether to engage with emerging platforms, but how to optimize timing, resource allocation, and competitive positioning within an inevitable participation framework. Understanding cycle progression enables proactive strategy development rather than reactive competitive responses."

How Do You Play The Game?

As I said on Unsolicited Feedback, my intention is not to paint a picture that these platforms are “good” or “evil.” I’m passing no judgement. This cycle happens do to competitive and capitalistic incentives.

My personal mentality is that “it is what it is.” But it’s a game. They are playing you, so you need to know how to play them. Given that, the most common question I got on the original post was - So what do I do next?

What’s Your Betting Strategy?

While my personal prediction centers on ChatGPT achieving platform dominance, this outcome is not guaranteed. The current competitive environment mirrors historical distribution shifts where multiple viable candidates compete for market control before a winner emerges.

• OpenAI with ChatGPT

• Anthropic with Claude

• Google with Gemini

• Meta with Llama and Meta AI

• Apple with ??? (it will come eventually)

This competitive dynamic directly replicates previous distribution shifts. Facebook competed against MySpace, Orkut, Hi5, and Friendster before achieving dominance. Google emerged victorious despite Yahoo's early distribution supremacy. The pattern demonstrates that initial market position does not guarantee long-term platform control. That means at this stage you are taking bets and you need a betting strategy.

Diversify or YOLO?

The two most common reactions to this stage are:

1. I’ll be early, but diversify across platforms until a winner emerges.

2. I’ll just wait to dedicate any resources until it’s clear who the winner is.

These are rarely the right strategies. Waiting until the winner is clear means you are typically too late. The early arbitrage advantages are probably gone and you will be fighting an uphill battle.

Being early and diversifying might be possible for very large companies that have large experimental budgets to spread around. Startups don’t have this advantage. They have resource constraints and need to focus. You need to bet, and you need to bet right. Higher risk, higher reward. In other words, portfolio strategy does not apply to startups.

gave a couple of great examples in the Unsolicited Feedback Episode:

"In the mobile shift, if you bet on just Apple, you could be a winner, like Instagram. If you bet on just Android, you could not. If you did Apple and Android, you would grow faster, it turned out, because both those platforms are dominant. But building for both took a lot of resources at the time. And building on Windows Phone was a huge waste of time.”

He went on to give another example in the shift to Social:

“The same was true on the Facebook platform. Some people did ok just on MySpace in the early days, but all of those developers shifted to Facebook, right? And there were a million other networks and there was a lot of encouragement from investors, from thought leaders, et cetera, saying should build as much distribution across them. It turned out the winning move were those that just built on the Facebook platform."

The point Fareed is making, is that every ounce of energy you spend on building on the platform that is not the ultimate winner has a massive tradeoff to spending energy on the platform that does end up being the winner.

Evaluating Platforms As They Emerge

If concentrated betting represents the optimal strategy for resource-constrained organizations, the critical question becomes: how do you identify the winning platform before market consensus emerges? The answer lies in developing a systematic evaluation framework that prioritizes leading indicators over lagging metrics.

Most strategic decision-making begins with the wrong variables, leading to systematically flawed platform selection. Understanding the hierarchy of platform evaluation criteria enables proactive positioning rather than reactive competitive responses.

Scale (Necessary But Not Sufficient)

Scale represents the most intuitive starting point for platform evaluation. The logic appears sound: larger user bases provide greater distribution potential, more integration opportunities, and higher visibility for platform participants.

The Scale Requirement Reality

Platform viability does require meaningful scale asymmetry. The platform user base must significantly exceed the typical size of integrating applications or services. This threshold ensures sufficient distribution potential to justify integration investment and ongoing platform dependency.

The Scale Limitation

However, scale alone provides insufficient strategic guidance. Historical examples reveal a consistent pattern: the eventual winner is never the platform with the largest initial distribution. Facebook defeated MySpace despite having a smaller user base at the time the platform emerged. Google surpassed Yahoo despite inferior initial reach.

Scale is a necessary but insufficient condition for platform viability. Organizations that prioritize scale over other evaluation criteria systematically bet incorrectly.

Retention & Engagement (The Real Signal)

Retention and engagement quality represents the most important platform evaluation criterion because it is most predictive of competitive sustainability, user retention, and ecosystem development potential.

Retention & Engagement Creates Category Winners

Retention and engagement depth consistently determine category winners across every major platform transition. This occurs because engagement drives the fundamental loops that fuel platform growth: user acquisition, monetization optimization, and competitive differentiation.

The Engagement Compounding Effect

High-engagement platforms create compounding advantages through several mechanisms. Deep user engagement generates superior data collection, enabling better personalization and more accurate recommendations. Strong engagement patterns create higher switching costs (habits are hard to break).

User Quality (Monetization Potential)

Different platforms also have different monetization potential. The classic example involves iOS versus Android platform users, where iOS users demonstrated higher app purchase rates, in-app spending, and premium service adoption despite Android's larger global user base. Early in the platform wars, Facebook users monetized much more effectively than Myspace users.

Value Exchange (Understanding the Rules)

You also need to analyze the underlying value exchange mechanisms that govern platform participation and ecosystem development. Successful platforms establish clear value exchanges where participants receive capabilities, distribution, or other benefits in exchange for extending platform functionality, providing data, or contributing to ecosystem growth. These are the “rules of the game” that you are playing. You need to deeply understand the rules to play effectively.

Oddly, platforms that are behind will offer incentives to artificially boost your presence in order to attract new developers and contributors. But this value exchange is unsustainable. They can’t artificially boost you forever, and they can’t offer it to everyone.

Evaluate platforms based on sustainable organic distribution mechanisms rather than temporary promotional incentives. The platforms offering the most authentic, user-driven distribution mechanisms are better.

It's About How You Exit, Not Enter

Most companies obsess over platform entry timing while completely ignoring the more critical challenge: building sustainable competitive positioning before the platform inevitably closes its gates. You have to enter with the end in mind.

Platform cycle navigation requires mastering two distinct competencies: entering at optimal timing and utilizing platform resources to build independent defensibility before the closure period arrives. The first challenge demands market analysis and strategic timing. The second challenge demands execution excellence and architectural foresight.

Entry timing mistakes create missed opportunities. Exit preparation failures create existential threats. The companies that die become platform-dependent without building alternative value propositions.

The Path

Building sustainable competitive positioning during platform participation requires independent value creation that leverages platform resources while developing platform-independent competitive advantages. I believe this comes from four steps.

1. User Experience Control: The Foundation Layer

You must control meaningful portions of the user experience rather than existing purely as platform extensions. Context and memory accumulation stems from controlling part of the user experience.

User experience control enables direct relationships with end users, unmediated feedback loops for product development, and control over retention and engagement optimization. Companies that allow platforms to own complete user experiences become features rather than products, vulnerable to absorption or replacement.

2. Workflow Integration: The Embedding Strategy

Controlling part of the user experience comes from embedding into use case specific workflows that extend beyond general platform capabilities. This creates specialized value propositions that platforms cannot easily (or don’t bother to) replicate.

3. Specialized Context Accumulation: The Data Moat Development

Workflow integration creates the opportunity to collect use case-specific context and memory that improves through sustained usage.

4. Micro Data Network Effects: The Competitive Isolation Strategy

Specialized context accumulation enables a micro data network effect that operate independently from macro platform network effects. While platforms like ChatGPT develop broad network effects through general context and memory accumulation, sustainable companies build specialized network effects.

The Race: Distribution vs Innovation

When new distribution windows open they create extraordinary opportunities, extraordinary casualties, and lots of missed opportunities. Success depends on winning a single, decisive competitive race that Alex Rampell captured perfectly in his 2015 blog post Distribution vs Innovation:

"The battle between every startup and incumbent comes down to whether the startup can get distribution before the incumbent can build the innovation."

This is not a fight for slow movers.

Startups possess innovation advantages through architectural freedom and focused development. Incumbents wield distribution weapons through existing user relationships and resource scale. Platform transition periods create compressed windows for market positioning.

If you're a startup:

1. Pick your platform bet and commit fully. Half-measures lead to irrelevance.

2. Enter with an exit strategy. Build independent defensibility while leveraging platform resources.

3. Move fast. The innovation replication cycle is brutal.

If you're an incumbent:

1. Don't underestimate distribution-focused competitors during platform transitions.

2. Focus on user experience and workflow integration, not just feature parity.

3. Build your own platform relationships before competitors do.

Case Study: Cursor vs Github Copilot

The GitHub Copilot versus Cursor market dynamic exposes the harsh mathematics of this competitive race. GitHub Copilot collapsed from 100% market share to 45% in twelve months as shown in this Ramp data tweeted by Alex Immerman.

Twelve months. GitHub had almost all the Advantages:

• Microsoft ecosystem integration across VS Code

• Established developer relationships spanning millions of users

• Massive resource allocation capabilities

• Preferred relationship with OpenAI through Microsoft

• Brand recognition and enterprise sales infrastructure

None of it seems to be mattering.

Distribution velocity focused on end-user experience demolishes structural advantages when executed with precision during platform transitions.

Achieving Escape Velocity

In The Great AI Rebundling, said:

But right now, we're witnessing something unprecedented: the fastest, most aggressive rebundling wave in tech history.

Modern incumbents are replicating AI innovations within months, not years. Large language model APIs democratize core capabilities. Cloud infrastructure enables instant scaling. Historically we saw a 24+ month innovation lead time but that has compressed to what seems like 3-6 months.

This creates a new dynamic for startups. You must achieve escape velocity in a much shorter window. This requires to throw away some of the traditional thinking of starting small and iterating slowly and methodically.

Strategic Reality Check

A lot of companies will lose this transition. They will make the wrong platform bets. They will diversify when they should focus. They will enter platforms without exit strategies.

But there will be a lot of new companies that are created in this transition.

Winners choose their platform bet and commit completely. They understand that half-measures lead to irrelevance.

Winners will own their user experience. They control meaningful user touchpoints. They embed into specialized workflows. They accumulate specialized context and data. They build micro network effects independent of platform control.

Winners achieve escape velocity faster than incumbents can copy. They will build product-led growth engines that will bypass enterprise friction.

Winners will enter with the exit in mind. They will extract maximum value during open gate periods while building independent competitive moats. They understand that platform dependency is strategic suicide.

Your Move

The distribution shift is here. New platforms are being created. They will begin opening their gates. The innovation replication cycle has accelerated beyond historical precedent.

Your move.

Note: Since publishing this post, my prediction has started to play out. ChatGPT launched the early version of their App SDK and is planning on a bigger launch sometime in late 2025/early 2026. To know what to do next, I suggest reading this post: How To Navigate The AI Distribution Shift.

The Missing Piece of the AI Revolution

Here's something that should keep every product leader up at night: We're living through one of the most significant technology shifts in history, yet we're still distributing products like it's 2015.

The AI revolution has transformed how we build products. We can now create experiences that would have been impossible just two years ago. But there's a glaring problem—this technology shift has come without a distribution shift. In fact, AI is actively destroying the distribution channels we've relied on for decades. SEO traffic is plummeting as users shift to answer engine platforms. Social algorithms are more unpredictable than ever. The old playbooks are burning.

But I believe we're on the precipice of something bigger. The next great distribution shift is coming, and it's going to reshape how products find users just as dramatically as search engines and app stores did before it.

When this shift happens—and the early signals suggest it's already beginning—there will be extraordinary windows of opportunity. There will also be casualties.

Every major new platform follows the same playbook: They start open and generous, practically begging developers to build on their platform. They need you to help solidify their moat. But once that moat has escape velocity? The walls go up. The taxes increase. The rules change. What was once free becomes paid. What was once permitted becomes forbidden.

Some platforms are surgical about it—Google took years to tighten its grip. Others are brutal—Facebook famously left "a trail of dead apps and battered businesses" when it shut down viral channels overnight. But they all follow the pattern.

The difference between the companies that thrive and those that get crushed? They understand the game being played. They see the pattern. They prepare for the inevitable shift from open to closed.

I first discussed the above concept with and on Unsolicited Feedback. Here is the full episode: YouTube :: Apple :: Spotify

In this post, I go much deeper:

• I'm going to show you exactly how this cycle works.

• We'll walk through the rise and fall of multiple major distribution platforms from Facebook to Google to LinkedIn.

• I’ll predict who I think will be the next distribution platform and why.

• Most importantly, we'll cover what you need to do now—before the window closes—to position your product for success in the AI platform era.

Because here's the truth: You can't opt out of this game. Your competitors won't. Your users won't. The only choice is whether you play it with your eyes open or closed.

Let's make sure you see what's coming.

Technology Shifts vs Distribution Shifts

In December 2023, published an important essay that I’ve been thinking about since. In "On Platform Shifts and AI," he made a distinction that every product leader needs to understand.

What separates a major platform shift from a minor one isn't the technology itself. It's whether that shift enables both new ways of building things AND new ways of reaching people.

These dual shifts—technology plus distribution—are what create the conditions for new multi-billion dollar companies to emerge. But here's where Winters' insight gets really interesting, and where most people miss the pattern:

"What I realized having gone through the internet and mobile platform shifts is that the technological and distribution shifts did not happen at the same time. Platform shifts that create both technological and distribution opportunities happen in a sequence, not all at once."

This sequence is everything.

As said:

The internet created websites in the early 1990s, but Google didn't emerge as the dominant distribution mechanism until 1998—and didn't truly monetize until the 2000s. That's nearly a decade between the technology and the distribution.

Mobile apps launched with the iPhone in 2008. Everyone rushed to build apps, convinced the App Store would be the distribution revolution. Remember "there's an app for that"? But most of those early apps died. The real distribution shift came four years later when Facebook launched mobile ads in 2012. Suddenly, apps could find users at scale, and companies like King (Candy Crush) and Supercell (Clash of Clans) exploded from nothing to billions.

Which brings us to AI.

We're clearly in the middle of a massive technological shift. AI has enabled entirely new ways of building products—from conversational interfaces to autonomous agents to personalized experiences that adapt in real-time. The technology is here, and it's spectacular.

But where's the distribution?

Yes, we’ve had break out companies like Cursor that have reached immense scale and broken growth records. But this has been off the back of word of mouth driven by the technology, not a new sustainable distribution channel.

ChatGPT has 700 million users, yet it's not a distribution channel, it's a destination. If anything, AI is destroying distribution channels. SEO is collapsing as answer engines like Perplexity, ChatGPT Search, and others keep users in its experience rather than sending traffic out. Social platforms are becoming walled gardens, limiting traffic being sent elsewhere. X has banned links, LinkedIn punishes posts that link externally, and the list goes on. The old channels are dying, but the new ones haven't emerged.

As Casey Winters put it in his December 2023 essay:

"ChatGPT is 'not it,' as the kids would say. At least not yet."

But, and this is crucial, we shouldn't expect the distribution shift to have happened yet. If history is our guide, we're right on schedule. The AI technology shift started gaining mainstream adoption in late 2022 with ChatGPT. If it follows the mobile timeline, we wouldn't expect to see the distribution shift until about 2026 or 2027.

Except I think it's going to happen faster this time. The platforms have learned from history. They know the playbook. And the early signals suggest they're already moving.

The question isn't whether a new distribution channel will emerge. It's who will control it, how open it will be, and whether you'll be ready when it arrives.

The Repeating Cycle: From Open To Closed

I believe we're at an inflection point. The distribution shift that's been missing from the AI revolution is about to arrive.

But before I explain who I think it is going to be and why, you need to understand the playbook. Because every major distribution platform in history has followed the same three-step cycle. Once you see it, you can't unsee it.

The Three-Step Cycle

Step 1: Identify the Moat

Every platform starts by figuring out what will make them unassailable. For Facebook, it was the social graph—who knows whom. For Google, it was search data—what people want. For Apple, it was having an application ecosystem to attract premium device owners. The moat determines everything that follows.

Step 2: Open the Gates

Once they know what they need, platforms become generous. They create "open" ecosystems, practically begging developers to build on top of them. Free API access. Viral growth mechanics. Revenue sharing that seems too good to be true.

Why? Because they need you. Your apps, your content, your data, your innovations—they all feed the moat. Every developer or creator who builds on the platform makes it stronger and harder to displace.

Step 3: Close for Monetization

This is where the pain happens. Once the platform achieves escape velocity, once the moat is so strong that users have no choice but to stay, the rules change. What was free becomes paid. What was permitted becomes restricted. The platform starts competing with its own developers, sometimes killing the very businesses that helped it grow. Sometimes this is deliberate and planned, sometimes this is a result of an emergent business need. But the ending is always the same. They do this systemically in three ways:

1. Build first-party versions of the most popular third-party applications.

2. Tax directly by taking a percentage of revenue.

3. Tax indirectly by artificially reducing organic reach to push towards their paid advertising.

Critical Warnings About This Cycle

1. The timeline is unpredictable but accelerating. Facebook went from open to closed in about two years. Apple took four. Google stretched it over two decades. You can't set your watch by it, but you can watch for the signals. I believe in AI, the players will be more ruthless because the stakes are much larger.

2. Step 3 is a bloodbath. When platforms close, there are casualties. Facebook's 2010 policy changes caused Zynga's player count to plummet overnight. Apple's in-app purchase requirements destroyed entire business models. Most developers are completely unprepared for how aggressive the platform will become.

3. Knowledge is survival. The companies that thrive through platform shifts aren't necessarily the biggest or best-funded. They're the ones who understand the game being played. They build with the end in mind. They extract value during Step 2 while preparing for Step 3.

The platform isn't your friend. It's not your enemy either. It's a force of nature. You can surf the wave or get crushed by it, but you can't stop it from coming.

Let's look at exactly how this cycle has played out across major distribution channels. The patterns are so consistent it's almost eerie.

Facebook Developer Platform

In May 2007, Facebook wasn't the inevitable winner we know today. They had 25 million users—impressive, but hardly dominant.

The competitive landscape was brutal:

• MySpace: 107 million users (over 4x Facebook's size)

• hi5: 70 million users

• Friendster: 50 million users

• Orkut: 30 million users

• Plus regional champions like Bebo (UK), StudiVZ (Germany), Vkontakte (Russia) and Tuenti (Spain)

Facebook was just one player in a crowded field, and there was no guarantee they'd win. They needed something to hit escape velocity before another platform locked in the global social graph.

Step 1: Identifying the Moat

Facebook recognized their moat was the social graph itself—the network of who knows whom. It's a classic direct network effect: the platform becomes exponentially more valuable as more of your friends join. But they had a problem. Building that graph user by user, friend by friend, was too slow. They needed acceleration.

Step 2: Opening the Gates

In May 2007, Facebook launched f8 and their developer platform. Mark Zuckerberg's opening line at the announcement is hilarious in hindsight:

"Until now, social networks have been closed platforms. Today, we're going to end that."

The irony of that statement would only become clear years later. Facebook's initial offer to developers was remarkably generous:

• We’ll provide you a canvas to build anything you want on our platform

• Use our viral channels to grow as fast as you can

• Keep 100% of your app's revenue

• We'll just monetize the sidebar ads

The response was explosive. By late 2007, over 7,000 apps existed on the platform, with 100 new apps launching every single day. By mid-2008, there were 33,000 apps and 400,000 registered developers. Users couldn't install apps fast enough—games, quizzes, vampire bites, zombie attacks, you name it.

The strategy worked perfectly. Each app helped bring in new users, drive engagement, and expand the graph. Facebook grew from 25 million to 250 million users in just two years, leaving MySpace and the others in the dust.

Step 3: Closing for Control

Once Facebook achieved escape velocity, the generosity ended. The closing happened in waves, each more restrictive than the last.

2009-2010: The Viral Channels Disappear

• Notification limits introduced

• App-to-user communication restricted

• News feed algorithm changes reduce app visibility

• Invite systems throttled

Zynga, the social gaming giant behind FarmVille, reported that after the 2010 policy changes, "the number of our players on Facebook declined." This was a company valued at $10 billion, brought to its knees by a few API changes.

2010-2011: The Tax Man Comes

• Facebook Credits becomes mandatory for all apps

• Facebook takes 30% of all transactions

• Direct payment methods banned

• Alternative monetization restricted

2011-2012: The Feature Absorption

• Facebook launches native alternatives to popular apps like Photos, Events, Groups, Location, Marketplace, and more.

• Platform features that apps relied on get deprecated

• Custom page tabs removed (BandPage lost 90% of its traffic overnight, dropping from 35 million to 3 million users)

By 2012, what Zuckerberg had proclaimed as an "open platform" had become one of the most tightly controlled ecosystems in tech. Facebook's platform left behind a trail of dead apps and battered businesses.

The most telling detail? Facebook insiders jokingly called their developer relations effort "Operation Developer Love"—acknowledging that it was more about extraction than partnership. Brutal.

It wasn’t just apps as well as Casey Winters notes:

Facebook pages worked the same way. They incentivized companies to grow page likes to get free distribution. Then distribution turned paid even if a user liked your page.

The Lesson

Facebook's platform shift established the modern template:

1. Use developers to build what you can't build alone

2. Let them take the risks and validate use cases

3. Once you've won, take back control and monetize aggressively

Every platform has run this same play. The only variables are timing and temperament. Some platforms are patient predators, waiting years before tightening the noose. Others move fast and break things—including their developers' businesses. Some are doing it on purpose. Some are doing it as a reaction to a business need. It does not matter. It always ends the same. Let’s take a look at a few more examples.

Apple App Store

Just as Facebook faced fierce competition in social networking, Apple entered the smartphone market as an underdog. In 2008, when the App Store launched, the mobile landscape looked nothing like today's iOS-Android duopoly:

• Nokia: 40% global market share with Symbian OS

• BlackBerry: The enterprise standard, growing rapidly

• Windows Mobile: Microsoft's strong presence

• Android: Just launching, but backed by Google

The iPhone was revolutionary but expensive, limited to AT&T, and lacked basic features like copy-paste. Apple needed something to make the iPhone indispensable—to justify the premium price and lock in users before Android gained momentum.

Step 1: Identifying the Moat

Apple recognized their moat wasn't just hardware—it was the ecosystem. If they could make the iPhone the platform with the best apps, apps you couldn't get anywhere else, they'd create switching costs so high that users would never leave. The network effect was more apps led to more iPhone users which led to more apps. But Apple couldn't build all those apps themselves.

Step 2: Opening the Gates

In July 2008, Apple launched the App Store with what seemed like a developer-friendly proposition at the time:

• 70/30 revenue split (developers keep 70%)

• Free app hosting and distribution

• Seamless payment processing

• Access to millions of credit cards on file

• Global reach from day one

Steve Jobs pitched it as revolutionary: "This is the best deal going to distribute applications to millions of users."

The results were staggering. "There's an app for that" became a cultural phenomenon. By 2010, the App Store had 225,000 apps and had generated $2.5 billion in revenue. Developers were making real money—some becoming millionaires overnight with simple apps like iFart and Beer Pong.

The App Store became the iPhone's killer feature. While Android was free, iOS had the apps people wanted. The moat was established.

Step 3: Closing for Control

Unlike Facebook's relatively quick closure, Apple played a longer game. The restrictions came slowly, each justified by "user experience" or "security," but the pattern was unmistakable:

2011: The In-App Purchase Mandate

• All digital content must use Apple's payment system

• No links to external payment methods allowed

• 30% commission on all digital goods

• Physical goods exempt (for now)

Amazon's Kindle app, which had allowed in-app book purchases, was forced to remove its store entirely. Netflix, Spotify, and others scrambled to adapt. The message was clear: pay the tax or lose access to iOS users.

2012-2015: The Slow Squeeze

• Restrictions on app updates and pricing changes

• Bans on apps that "duplicate core functionality"

• Rejection of apps that compete too directly with Apple

• Introduction of Search Ads in the App Store reducing organic distribution and visibility

2016-Present: Maximum Control

• Apple systematically builds features that kill popular apps

• App Tracking Transparency decimates ad-based business models

• App Store Small Business Program (15% commission) seems generous but comes with restrictions

• Battles with Epic Games, Spotify, and regulators worldwide

• Alternative app stores and side loading still forbidden

Today, Apple's control is absolute. Apple has made departure impossible. Want to reach iPhone users? You go through Apple. Period. The 30% tax that seemed reasonable in 2008 is now generating over $100 billion annually.

Google Search

Google's platform shift is unique—it took two decades to fully play out. While Facebook and Apple showed their cards relatively quickly, Google maintained the illusion of an open web for much longer.

In the early 2000s, search was fragmented. Yahoo served as the starting point for millions. AltaVista appealed to engineers. Microsoft pushed MSN Search. Dozens of specialized engines fought for relevance. Google was just another upstart with a clean interface and remarkably better results.

Step 1: Identifying the Moat

Google's moat was elegantly circular: data and content. The more people searched, the more data they gathered. The better their results, the more people searched. But this virtuous circle had a critical dependency—they needed the entire web's content.

Unlike Facebook or Apple, Google's moat required universal participation. They needed millions of websites creating content, structured in ways their crawler could understand. The open web wasn't just philosophy—it was a business requirement.

Step 2: Opening the Gates

Google's early philosophy bordered on religious devotion to openness. Larry Page captured it perfectly in 2004:

"We want to get you out of Google and to the right place as fast as possible."

This wasn't just rhetoric. It was the value exchange that built the modern web:

• Create good content → Google sends free traffic

• Optimize for their algorithm → Rank higher

• Better content → More visitors

This spawned the entire SEO industry. Businesses were built on ranking #1 for valuable keywords. Publishers reached audiences without paying for distribution. Small sites could outrank giants with superior content.

For over a decade, it felt like genuine partnership. Google's revenue came from clearly marked ads in yellow boxes or sidebars. The main results remained sacred—organic, merit-based, untouchable. Ranking #1 organically was the holy grail of internet business.

Step 3: Closing for Control

Google's closure happened so gradually it felt like evolution, not strategy.

Phase 1 (2010-2015): The Ad Creep Ads moved from sidebar to prime real estate above organic results. Three to four ads became standard for commercial queries. Google killed keyword data with "not provided," blinding websites to their traffic sources.

Phase 2 (2015-2020): The Feature Takeover Featured snippets answered questions without clicks. Knowledge Graph pulled data without attribution. "People Also Ask" pushed organic results down. Local packs, shopping results, flight information—all Google products claiming more territory.

Phase 3 (2020-Present): The Closed Garden The Markup's 2020 investigation revealed the stunning reality:

• 41% of mobile first screen shows Google products

• 63% of visible first page is Google-controlled

• 1 in 5 searches show zero external results initially

Today, a commercial search reveals Google's true form. You'll see ads, then shopping results, then maps, then featured snippets, then "People Also Ask" boxes. Actual organic results? Maybe on scroll eight.

Google began competing directly in lucrative verticals. Travel queries that fed Expedia now show Google Flights. Product searches bypass Amazon for Google Shopping. Local searches favor Google Maps. Companies that spent decades building SEO-driven businesses watched their traffic evaporate—not because they got worse, but because Google decided to compete.

Google has essentially become a closed distribution ecosystem atop the open web: it still indexes the web, but it increasingly interposes its own monetized layers (ads) and products. SEO is no longer a guarantee of reach. Even if you rank well, you might be shown below a block of sponsored results and Google’s own info panels.

LinkedIn

LinkedIn's platform shift is happening right now, in real-time. It's also the fastest we've seen—from open to closed in less than four years. If you're a business content creator who's noticed your engagement plummeting, you're not imagining it. You're witnessing the playbook in action.

For most of its existence, LinkedIn was essentially a digital resume database. Users updated their profiles maybe once a year. Engagement meant accepting connection requests. But there was an internal mandate: transform LinkedIn from a static directory into a daily destination.

Step 1: Identifying the Moat

LinkedIn's moat is professional data—not just who you are, but what you do, what you care about, and how you behave professionally. This data is gold to three lucrative customer segments: marketers targeting B2B buyers, recruiters hunting talent, and salespeople seeking warm leads.

But static profiles only reveal so much. LinkedIn needed dynamic data—what you read, what you share, what makes you engage. They needed users actively participating, not just existing.

Step 2: Opening the Gates

Around 2020, LinkedIn made a strategic decision to prioritize "creator" content in the feed. The deal was irresistible: publish business content directly on LinkedIn, and they'd give you extraordinary organic reach.

The early results were staggering. Posts that would get 100 views on other platforms were getting 10,000+ impressions on LinkedIn. Some creators went from unknown to industry thought leaders in months. The algorithm seemed to love everything—text posts, native video, document carousels, polls.

LinkedIn sweetened the deal by launching Creator Mode, newsletter functionality, and live video. They actively courted business influencers to post exclusively on the platform. The message was clear: this is your new distribution channel.

For about two years, it was a content heaven. B2B marketers abandoned Twitter and blogs for LinkedIn posts. Sales teams trained executives to become "thought leaders." Everyone rushed to claim their piece of the organic reach pie.

Step 3: Closing for Control

The first signs of closure appeared in late 2022. Creators started noticing declining reach. Posts that once reached 50,000 professionals now struggled to hit 5,000. The algorithm became pickier, favoring certain formats while punishing others.

Then came the smoking gun: Thought Leader Ads.

Launched in June 2023, this new ad format allows companies to sponsor organic posts from individuals—turning executives and influencers into ad inventory. The implication was obvious: if you want reach, you'll need to soon start to pay for it.

The data confirms what creators feel. Social media agencies report LinkedIn organic engagement rates hitting all-time lows, with algorithm changes causing 50-75% decreases in reach for posts that previously performed well. The same content, the same creators, dramatically different results.

Today, LinkedIn's feed is starting to look more like driving down the 101. Sponsored content dominates. Organic posts are getting buried under "suggested posts" from people you don't follow. The creator economy LinkedIn built is being systematically monetized.

LinkedIn is a proof point that platform cycles are accelerating. What took Facebook five years and Google twenty years, LinkedIn accomplished in less than four. They've perfected the formula:

1. Create artificial scarcity: Make organic reach feel special and exclusive

2. Build creator dependence: Encourage people to abandon other channels

3. Flip the switch: Once critical mass is achieved, monetize aggressively

4. Gaslight the community: Claim the algorithm changes are about "quality"

The Impending New Distribution Channel: ChatGPT

By now, the pattern should be clear. Every major platform follows the same playbook: open, grow, close, monetize. My personal belief is that ChatGPT is the next distribution platform. I think it will happen within the next 6 months (and possibly faster). Even if it’s not ChatGPT, it will be someone else and the pattern will repeat.

Step 0: The Competitive Environment

Despite ChatGPT's 700 million users, we're still in a genuinely competitive environment. The AI model leaderboard changes monthly. OpenAI releases GPT-4, then Anthropic counters with Claude. Google launches Gemini. Meta open-sources Llama. DeepSeek emerges from nowhere with competitive benchmarks.

Users still have relatively low switching costs. Copy your prompt, paste it elsewhere, get similar results. There's no lock-in yet. No data gravity. No compelling reason to stay with one provider.

Some of you are probably screaming at the monitor about Claude, Google Gemini, Meta, and Apple as being in the market. But remember, Facebook wasn’t the biggest social network. Apple didn’t have the most popular phone. Google was surrounded by other search engines.

This is what this stage of the cycle always look like—before someone figures out the moat.

Step 1: Identifying the Moat

OpenAI has identified their moat, and it's not model quality. It's context and memory.

Think about what makes an AI assistant truly valuable. It's not just answering questions—it's understanding your specific situation, remembering your past conversations, knowing your preferences, and building on previous interactions. The AI that knows you've been working on a product launch for three months and can reference your previous discussions is infinitely more valuable than one starting from scratch each time.

From Aaron White (Founder of, Former CTO at Vendr) on a recent Unsolicited Feedback podcast episode:

“When AI is fully ingesting and memorizing everything you see and have seen it can begin promoting information to you w/o you even prompting. That requires long term, massive memory- and it’s clear between baby-memory, MCP, and their desktop app which has hooks to “see what you see” they are heading down this path.”

The key here is that MCP only addresses context portability, not memory. The accumulated history of interactions, the learned preferences, the refined understanding of each user—that's not portable. And that's where the real lock-in lives. OpenAI seems to be moving fastest on this.

This is different from previous platform moats. Facebook needed your social graph. Google needed the web's content. OpenAI needs something far more intimate: your entire digital context. Every document, every conversation, every preference, every workflow.

Step 2: Opening the Gates

They can't capture this context and memory alone. No single company can build integrations with every tool, access every data source, or understand every workflow. So we're seeing the early signs of platform opening.

There are a few early signals that they are about to open the gates:

• They recently launched connectors w/ Deep Research to tools like HubSpot, Box, GitHub, and more.

• OpenAI is actively soliciting more platform integrations.

• They're hiring aggressively for their Agent Platform/Infrastructure team, including “Product Manager, Agent Platform” with job postings explicitly stating they're building "the platform and integrations to launch new agents to hundreds of millions of users worldwide."

Each of these integrations and agents will provide more context to create more usage and enable OpenAI to capture more memory increasing the moat. We’ll have to see how the value exchange between OpenAI and developers will be constructed.

Remember, they’ve already attempted this twice. First with GPT store promising monetization was coming (but never did). Then with plugins, which they shut down. They are looking for the right formula.

Step 3: The Inevitable Closure

Once OpenAI achieves context lock-in, when switching to another AI means losing months or years of accumulated context, the platform dynamics will likely shift.

The first changes will be subtle. Maybe a "ChatGPT Platform Fee" for high-volume integrations. Perhaps preferred placement for "certified partners" who pay. Possibly restrictions on what types of data can be accessed or how it can be used. Or a platform tax like Apple leverages on the mobile ecosystem.

Then comes the feature absorption. The most successful use cases built on the platform will be "simplified" into native ChatGPT features. They’ve already done this around use cases for critical context (i.e. Glean and Granola). Why use a third-party research tool when ChatGPT Pro includes it? Why maintain a separate workflow automation when ChatGPT Actions handle it natively?

The worst-case scenario would be a complete platform shutdown. They have the context, the users, the habit. They pull a Facebook: revoke API access (or Twitter - see Meerkat), build everything in-house, leave a trail of dead startups in their wake.

Are these platforms “evil?”

I want to be clear. I’m not implying OpenAI is “evil” or deliberately planning this from the start. It’s just the reality. The cycle always ends the same.

The cycle emerges from natural competitive dynamics within capitalism. Companies exist to grow. To grow they need to gain a competitive advantage. This competitive environment in capitalism creates an incredible amount of consumer value. Time and energy spent thinking about how this cycle may be unfair is misdirected. You should re-route that energy to understanding the rules of the game and how to participate.

Why OpenAI / ChatGPT Over Others?

OpenAI is showing many signals they are about to make this move. Claude, Google, and others could make a similar move. So why is my prediction on OpenAI?

If you are a Reforge alum, you should have one thing burned into your brain - those with the highest retention & engagement win categories. Retention is the god metric. It is a universal law at this point and has been proven over and over.

ChatGPT clearly has the highest retention. Deedy Das, VC at Menlo Ventures recently posted this data. What do you see?

I see three things:

1. They have the highest mean customer retention vs others in the category.

2. Their retention curves have been shifting up over time.

3. They have attained the rare and magic “smile curve” of retention.

The only other times I have seen this pattern is in retention curves of products like Facebook, LinkedIn, and Slack. I’ll repeat, it does not matter that Google / Meta have access to larger distribution if the retention & engagement disparity is this large.

It will fuel the engagement loops that will drive integrating more context and capturing more memory. It will also attract more developers and builders to their inevitable platform fueling it even more.

The Accelerating Timeline

Casey Winters observed something:

"With every generation, companies that reach massive scale have gotten more efficient at preventing other companies from growing on top of them, at least for free."

Google let SEO flourish for two decades. Facebook gave developers five years. LinkedIn managed less than four. How long will the AI window last? My bet: two years. Maybe less. They've learned from history. They know the playbook. They'll move faster than anyone expects. The stakes are much larger.

The integration requests, the MCP protocol, the agent platform—these aren't just features. They're the early moves in a platform play that will reshape software distribution. OpenAI isn't building an AI assistant. They're building the next destination, the next App Store, the next search engine, the next social network—except this time, the moat isn't a social graph, user generated content, or connections. It's context and memory. And once they have yours, switching will become almost impossible.

There Is No Opt’ing Out

Here's the uncomfortable truth: knowing the game doesn't mean you can opt out.

Every product leader faces the same dilemma. In isolation, integrating with ChatGPT makes no sense. Why would HubSpot want to become a database to ChatGPT's interface? Why would Notion want to feed their replacement? Why would any company voluntarily feed the next platform monopoly?

But we don't operate in isolation. We operate in competitive markets. And markets are ruthless about efficiency.

When your competitor integrates with ChatGPT and their users love the AI-powered experience, what choice do you have? When customers start asking why your product doesn't work with ChatGPT like your rival's does, what's your response? When the trade press writes about how you're falling behind in the AI revolution, what's your strategy?

This is the prisoner's dilemma of platform shifts. Everyone knows how it ends, but everyone still plays because the alternative—letting competitors claim the early advantage—is worse.

The window is opening now. ChatGPT will soon cross a billion monthly active users. The competitive environment has peaked. The early integrations will get preferential treatment, better placement, and deeper access. They'll shape the platform rules. They'll build the habits. They'll capture the value before the taxes arrive.

But rushing in blindly is how you end up in the platform graveyard. The smart players are already asking the critical questions:

1. How long will the window stay open? Plan accordingly.

2. What's the real value exchange? It's not just API access for distribution. It's your data, your user relationships, your competitive advantage.

3. How aggressively will OpenAI monetize? Casey Winters notes that platforms get more efficient at value extraction with each generation. Expect OpenAI to be the most efficient yet.

4. Most importantly: How do you use the platform while building your own moat? This is where product strategy matters. You need to create value that persists even when platform access gets restricted. Build direct relationships. Capture first-party data. Create moats within your own product.

The companies that survived previous platform shifts didn't just integrate—they integrated strategically. They used Facebook's viral channels while building email lists. They leveraged Google's traffic while developing brand loyalty. They sold through the App Store while creating web experiences.

This time won't be different. Use ChatGPT's distribution, but don't depend on it. Let their AI enhance your product, but don't let it become your product. Most importantly, assume the platform will turn hostile and plan accordingly.

The stampede is coming. If I’m right, in six months, it’s likely every SaaS product and consumer application will be rushing to complete a ChatGPT integration. In twelve months, users will expect it. In eighteen months, the platform taxes will arrive. In twenty-four months, the graveyard will be full.

If you’ve gotten to this point in the essay, you can see it coming. You understand the pattern. You know the game. The only question left is whether you'll play it with your eyes open or closed.

Because in platform shifts, there are only three types of companies: those who move too early and waste resources, those who move too late and miss the window, and those who see the shift coming and position themselves to win. If you end up in the graveyard, it’s not ChatGPT’s (or whoever is the platform) fault. It’s yours.

The clock isn't just ticking, it's accelerating. And in this game, the house always wins.

But if you play it right, you can win too—at least for a while.

Choose wisely. Move deliberately. And whatever you do, don't be naive.

The next great distribution shift isn't coming. It's here. And whether you're ready or not, you're already playing.

Appendix: How This Prediction Could Be Wrong

Let me be clear about what I'm predicting here. I give it 80% odds that ChatGPT/OpenAI becomes the next major distribution platform. But I'm 100% certain that whoever wins will follow the same open-grow-close pattern we've documented. The cycle is too proven, too profitable to abandon.

Still, predictions are humbling, and there are legitimate reasons why ChatGPT might not be the one. Let's examine them honestly.

Apple Gets Their Act Together

Apple is the sleeping giant that could change everything. They own the device layer—both mobile and desktop. They see every interaction, every app, every moment of your digital life. They have the ultimate context and could build the ultimate memory.

As said on Unsolicited Feedback:

“If you own the device, you see everything the user sees — that’s why Apple is in the best position.”

If Apple suddenly shipped a compelling AI assistant that leveraged all that context, integrated with all their services, and opened it to developers, they could leapfrog everyone. They have the users, the developer relationships, and the platform experience.

But "if" is doing heavy lifting here. Apple's AI efforts have been underwhelming (and that is putting it politely). Their culture prioritizes privacy over the data collection that powers AI. Their famous secrecy conflicts with the developer openness platforms require.

OpenAI clearly sees this threat—hence their reported device collaboration with the acquisition of Jony Ive. They're not waiting for Apple to wake up.

OpenAI Fumbles It

This is the most likely failure mode. OpenAI could absolutely blow their advantage by moving too aggressively, too fast. They're already pushing boundaries competing with enterprise players like Glean, and launching copycat features like meeting recording.

If they get too greedy too quickly, they could create an opportunity for one of the competing players to come in with a more friendly platform. But remember, ChatGPT has the user advantage at 700M MAU’s and growing.

OpenAI Already Has Escape Velocity And Doesn’t Need A Platform

One possibility is that OpenAI already has enough escape velocity on ChatGPT and doesn’t need to create a platform to hit it. That might be true in isolation, but once again we live in a competitive environment. The rational thing for Claude, Google, others to do in this case is to build a platform themselves in order to accelerate their growth and compete. As a result, I think ChatGPT/OpenAI launches their own just to cement victory and stave off the competitive move of others.

MCP Makes Context More Portable

Anthropic's Model Context Protocol was a clever defensive move. In theory, it commoditizes integrations—if every AI can access the same context through standardized protocols, no single platform can monopolize it. It's like making phone numbers portable between carriers.

But theory isn't reality. Developers and businesses are resource-constrained. They'll focus on the largest platform first, and ChatGPT has at least 20X Claude's user base and is growing faster. Even if every AI platform launched identical integration capabilities tomorrow, the rational move is to optimize for ChatGPT's 700 million users, not Claude's 30-40 million.

This creates a compounding advantage. More integrations make ChatGPT more valuable. More value attracts more users and deepens engagement. More users attract more integrations. We've seen this movie before.

Plus, MCP only addresses context portability, not memory. I’ll repeat - the accumulated history of interactions, the learned preferences, the refined understanding of each user—that's not portable. And that's where the real lock-in lives.

There Are Major Other Players With Lots of Distribution

Microsoft has Copilot embedded across Office. Google has Gemini integrated everywhere they plaster the icon. Apple has... well, they're trying. Couldn't any of these giants become the platform instead?

Remember: Facebook wasn't the largest social network when it launched its platform. The iPhone wasn't the market leader when the App Store arrived. What matters isn't size—it's engagement depth, platform readiness, and trajectory.

We’ve seen the retention numbers. My hypothesis is that if we could see the engagement numbers between platforms, ChatGPT has the deepest engagement. They're creating the kind of deep engagement that makes a platform valuable to developers. Microsoft users might encounter Copilot, Google users might encounter Gemini CTA’s, but do they build their day around it?

Depth beats breadth in platform dynamics. Facebook had far fewer users than Myspace, hi5 or Friendster, but their users were addicted. That addiction—that depth of engagement—is what developers and creators want to tap into.

Regulation

This is the true wild card. The regulatory environment around AI is volatile and unpredictable. Data privacy laws, AI safety regulations, antitrust enforcement—any of these could derail a platform play.

Europe might decide that accumulating user memory is fundamentally anticompetitive. The US political environment has been on an absolute rollercoaster the last couple of years.

Regulation could slow the platform shift, fragment it geographically, or even prevent it entirely. But betting against technology platforms finding ways around regulation is historically a losing proposition. They'll adapt, lobby, and innovate their way through.

OpenAI’s Land Grab on Unsolicited Feedback

I first discussed the above concept with and on Unsolicited Feedback. Here is the full episode: YouTube :: Apple :: Spotify

Every decade or so, a technological shift comes along that doesn't just change how we build products—it rewrites the entire playbook. The shift from on-premise to cloud wasn't merely a hosting change; it revolutionized every aspect of how we build, ship, and grow software products. AI represents a similar inflection point.

Our belief at Reforge is that the technology shift to AI will completely redefine product teams and in a lot of ways I think the ecosystem is underestimating not overestimating the impact it will have over the next decade.

To understand AI's transformative potential, let's examine how the cloud revolution reshaped product development. This isn't just a history lesson—it's a blueprint for understanding the scale of change ahead:

How On-Premise To Cloud Redefined Product Teams

🔄 Development Methodology

The dominant development methodology pre-cloud was Waterfall. Product teams operated like factory assembly lines: Large rigid six-month release cycles, exhaustive requirements documents, and testing phases that lasted longer than development itself. The shift to cloud broke these constraints. Teams could now ship code continuously, in smaller more iterative releases, lightning the need for exhaustive testing and requirements. It wasn’t just faster development, it was a completely different way to develop.

🧰 Tools

Cloud computing didn't just add new tools to our stack—it catalyzed an entirely new ecosystem of product development technology. Version control and collaboration tools like GitHub and GitLab, continuous integration and deployment tools, cloud development environments, containerization platforms, cloud-native security and compliance, and feature flags just to name a few for engineering.

For product managers, tools like JIRA were used to manage iterative development cycles. Product Analytics was fueled now that we could get real-time usage data. A/B testing and experimentation platforms were now possible.

💰Monetization Models

Pre-cloud the dominant monetization model was transactional. Massive upfront licenses where the customer would hopefully upgrade on some multi-year cycle. The shift to cloud the enabled subscription to become the dominant monetization model. This enabled lower barriers to entry, free trial models, and different dimensions of charging for features and usage. This wasn't merely a pricing change—it altered the relationship between software companies and their customers, creating incentives for continuous improvement and customer success.

📈 Growth Models and Channels

The shift to cloud also fundamentally changed how technology products could find and acquire customers. It enabled product led growth, freemium models, viral expansion and bottoms-up adoption within organizations. It also enabled leveraging new online growth channels like search engine optimization, paid search, and paid social.

🏆 Measures Of Success

Cloud transformed what we measured and how we thought about success. On-premise teams tracked license revenue and maintenance renewals. Cloud teams moved to metrics that were previously impossible to capture: daily active users, time-to-value, feature adoption rates, and real-time usage patterns. This shift from lagging to leading indicators enabled product teams to build and grow in a different way.

🛡️ Defensibility

Cloud changed how technology companies built competitive moats. In the on-premise world, complex installation requirements and high switching costs created artificial lock-in. Cloud enabled—and required—new forms of defensibility: network effects that grow stronger with each new user, data advantages that improved product value, and platform ecosystems that increase in value as more developers build on top.

🧠 Skillsets & Roles

The cloud transformation didn't just change what product teams built—it redefined who we needed to build it. Pre-cloud product managers were primarily requirement gatherers and project coordinators. Cloud demanded a new breed of PM: data-driven decision makers who could analyze user behavior, run A/B tests, and optimize based on real-time metrics. Additionally, specializations emerged like the Growth Product Manager as new growth models became viable. The shift wasn't just about adding new skills—it changed how we evaluated product talent.

👏 Team + Org Design

Pre-cloud organizations were built like assembly lines: development threw code over the wall to QA, who threw it over to operations, who somehow had to keep it running. Large specialized teams organized by function (Development, QA, Operations) where each phase of project was handed off sequentially.

The cloud era introduced cross-functional pods. Smaller, more autonomous units comprising product, design, and engineering working together toward shared outcomes. The impact rippled up to executive leadership. Traditional IT organizations had CIOs who managed infrastructure and systems. Cloud demanded technology leaders who understood product strategy, customer experience, and business models. Chief Product Officer roles and product management grew substantially and expanded scope to include growth, data, and other domains.

AI Native Product Teams

New Constraints, New Possibilities

Just as the shift to cloud was much more than new technology, it’s easier to see that AI will not be an incremental change but a fundamental reinvention of how we build and grow products. AI is shifting the constraints and possibilities around:

• What products and features we build

• How we build those products

• How we grow products

• The roles, teams, and org structures

The next generation of product teams will be trained as AI-native from day one. They will think, work, and build differently. They will contain new roles, org structures, and processes. It is impossible to predict exactly what this looks like 5+ years from now. But here are some of the things that AI is enabling in the short term.

The First Wave: How AI is Already Reshaping Product Teams

1. From Exploring A Couple Paths To Exploring Many Paths

2. From Idea → Doc to Idea → Prototype

3. Problem Expansion, Prioritization Reshuffle, Solution Reinvention

4. Redefinition Of Product Team Roles

5. Growth Model and Channel Changes

6. A Major Retooling Of Fragmented Product Stacks

7. Realignment In How Products Create and Capture Customer Value

From Exploring A Couple Paths To Exploring Many Paths

In current product development, the pressure to ship often forces teams into premature convergence. Most teams can only afford to seriously explore one or two solution paths before committing—usually the ones championed by the loudest voices in the room. This linear exploration creates three critical problems:

1. High-stakes decisions based on limited data

2. Solutions optimized for internal consensus rather than customer value

3. Innovative approaches killed by time constraints before they can prove themselves

But as Scott Belsky, CPO of Adobe points out AI not only helps accelerate product cycles but also give us more exploration cycles:

“what makes this technology truly distinctive from other advances is its reasoning and imaginative capabilities (not taste-based imagination, but boundless directed exploration). What this technology really gives us is MORE CYCLES - more cycles to explore”

This expanded exploration capacity transforms every aspect of product development. Teams can simultaneously test multiple interface designs, generate dozens of copy variations, prototype competing technical approaches, and validate different go-to-market strategies—all while maintaining the quality of each exploration.

What historically took months of sequential iteration can now happen in parallel, dramatically increasing both the quantity and quality of product decisions. AI isn't just accelerating our existing processes; it's fundamentally changing how we discover and validate product opportunities.

From Idea → Doc to Idea → Prototype

The Document Death Spiral

The traditional product development playbook starts with documentation. Whether it's a PRD, an Amazon-style press release, or a detailed user story, product teams spend countless hours crafting documents that attempt to capture our product vision. These documents then enter the document death spiral- endless cycle of reviews, debates, and revisions—often becoming more about internal alignment than customer value.

The Understanding Gap

The fundamental flaw in this approach isn't the documentation itself—it's the gap between written description and shared understanding. When a product manager writes "intuitive user experience" or "seamless integration," each stakeholder envisions something different. This misalignment creates what I call "work around the work": endless meetings, revision cycles, and debates that consume energy without moving the product forward.

Clarity With Prototypes

Prototypes cut through this ambiguity. A working prototype—even a rough one—creates clarity and alignment that no document can match. It transforms abstract discussions into concrete decisions, replacing "I think" with "I see." But until now, prototyping has been gated by technical constraints: either you waited for engineering resources or settled for limited mock-ups.

AI is changing this constraint. It is on a path to enable engineers to build functional prototypes in hours instead of weeks, while empowering non-technical team members to create interactive demonstrations without writing code. This shift from "documentation-first" to "prototype-first" development won’t just save time—it fundamentally improves product decisions by grounding them in tangible experiences rather than theoretical discussions.

Problem Expansion, Prioritization Reshuffle, Solution Reinvention

At it’s core, a product team’s job boils down to three essential responsibilities:

• Understand the Customer’s Problems

• Prioritize the Customer Problems To Solve

• Facilitate the Solutions to Those Problems

AI has a large impact on all three of these.

Problem Expansion

AI doesn't just help solve existing problems— it expands the realm of what problems are solvable. This means your problem space is no longer constrained by old limitations. More specifically, as Ravi Mehta, creator of our upcoming AI Strategy course and former product leader at Tinder, Facebook, and TripAdvisor states we move from being able to solve computation problems to learning problems.

*“In the Learning Era, we are no longer limited by what we can express in code. Given a sufficient amount of data — for example photos tagged with whether or not they contain a bird — an AI model can learn how to detect birds in nearly any photo.

Even more importantly, AI models can be stacked on top of each other and combined with computational approaches to make extraordinary products that can work in magical ways. The capability to detect a bird — or any other object — has led to image and video generation. Quickly, machine learning evolved from understanding images to making them.”*

Discovering and understanding the learning problems will take a mindset and methodology shift.

Prioritization Reshuffle

While there are many product prioritization frameworks out there, they almost all consider the four dimensions of feasibility, impact, risk, cost. AI introduces new factors into each element of prioritization.

• Feasibility - Solutions that were once too complex or technically impossible become attainable.

• Impact - The ability to personalize at scale means certain problems—like creating hyper-targeted user experiences—become more impactful than ever before.

• Risk - AI introduces new risk considerations, such as the potential for hallucination, bias, or misinformation.

• Cost - AI can both increase and decrease costs, depending on model usage, scale, and complexity.

Solution Reinvention

AI also enables entirely new solution spaces for problems we identify and prioritize. Just a few dimensions we are seeing this manifest:

• Adaptive vs. Static Solutions - AI enables solutions that evolve and improve through usage. Instead of building fixed features, we're creating learning systems that adapt to user behavior, context, and emerging patterns. A recommendation engine doesn't just follow rules—it discovers new patterns of user preference we never anticipated.

• Do The Work vs Enable The Work - Many software products today are tools that enable a user to create content and experiences. Canva, Notion, Google Docs, Gmail, etc. But AI is enabling just doing the work for the customer vs enabling it.

• Scalable Personalization & Dynamic Interfaces - Previously, personalization for most products meant simple if-then rules or basic user segments. AI enables hyper-personalization that scales: interfaces that adapt to individual working styles, content that reshapes itself based on comprehension levels, and workflows that optimize for each user's unique patterns. What was once a choice between standardization and customization becomes both simultaneously.

Redefinition Of Product Team Roles

Blurring Roles

AI is enabling product managers to write code, engineers to take on product management work, marketers to code landing pages, etc. The boundaries of roles is starting to blur. The expectations are fundamentally changing. This will eventually impact how we organize product teams.

The Magic Of Early Stage Teams

The impact of this is greater than some different role definitions. In early stage startups, there is something special that happens that feels like magic. Teams tend to ship a lot more with a lot less people. As a company grows you try to maintain that magic but it eventually disappears. AI can potentially maintain the magic of early stage startups you lose over time.

What feels like magic, isn’t magic at all. Early stage startups create the constraints and conditions for:

• Super tight feedback loops between the builders and the customers.

• These super tight feedback loops create “founder intuition.”

• Everyone on the team does a little bit of everything because they have to.

But over time all those things get extracted away from the builders.

• Tons of specialized roles (user researchers, data analysts, sales, success, etc) get introduced creating layers between the customers and the builders.

• This slows down the feedback loops between the builders and customers which slows down the team building intuition.

• Intuition is replaced by documentation, product reviews, sync meetings in order to keep things “aligned” and headed in the right direction.

• These things replace building time with time spent on the work behind the work.

The list goes on. But when I look at some of the changes that AI can have on how we build products and the boundaries of roles, a lot of it feels like returning or at least maintaining how great early stage teams operate.

Growth Model and Channel Changes

The cloud revolution gave birth to product-led growth, freemium models, and viral expansion. Now, AI is poised to trigger an equally dramatic transformation in how products find and engage customers. Early signals suggest we're entering an era where traditional growth playbooks may become obsolete.

Traditional Channels Show Cracks

We're witnessing the first tremors of change in established acquisition channels. SEO strategies that worked for decades are being disrupted by AI-powered search engines that bypass traditional content. Email marketing effectiveness is declining as AI assistants filter and prioritize messages. Even paid acquisition channels are showing vulnerability as users adopt AI interfaces that reshape how they discover and evaluate products.

The Rise of AI-First Distribution

The next wave of growth channels will likely be AI-native. As Aravind Srinivas, CEO of Perplexity AI, provocatively suggests, maybe we move to a world where AI agents—not humans—become the primary audience for product promotion. Imagine pitching your product not to end users, but to the AI assistants that increasingly guide their purchasing decisions.

This shift has profound implications:

• Discovery Mechanisms: Products will need to be discoverable and evaluatable by AI systems, requiring new approaches to product metadata and integration points

• Value Demonstration: Instead of emotional appeals and brand messaging, products may need to demonstrate value through quantifiable metrics that AI can assess

• Integration Points: Success may depend less on traditional marketing and more on building the right API hooks and AI-readable documentation

Product Development Implications

As we've long observed in Product Channel Fit theory, products must mold to their distribution channels—not vice versa. When AI becomes a primary distribution channel, it will force changes in:

• How we structure and expose product capabilities

• How we communicate product value propositions

• How we design onboarding and integration experiences

• How we measure and optimize for adoption

The winners in this new era won't be those who simply adapt their marketing to AI channels, but those who rebuild their products with AI distribution in mind from the ground up.

A Major Retooling Of Fragmented Product Stacks

Product Stack Fragmentation

Product teams have gone through a chaotic accumulation of purpose built tools As a result, today’s product stacks are layers of tools accumulated over time, each solving a specific need but never truly integrated. Feature flags in one tool, analytics in another, customer feedback in a third, and the list goes on. While this fragmentation was manageable in traditional development, it becomes a critical vulnerability in the AI era.

The Compounding Error Problem AI systems don't just struggle with fragmented tools—they fail exponentially because of them. As Dharmesh Shah, founder of HubSpot, explains:

"Let's say an agent needs to invoke the LLM a dozen times to accomplish a goal...Now mathematically, if each invocation has just a 95% success rate — or a 5% error rate, the success rate of the final result is 0.95 to the 12th power which is about 54%. So basically a coin toss. Half the time you'll get something right-ish and the other half you'd get something wrong-ish."

Error rate increases when working across fragmented systems. Each boundary between tools becomes another opportunity for errors to compound.

AI-Native Product Stack

AI-native product stacks will require a fundamental rethinking. The teams that solve this integration challenge first will gain a significant competitive advantage—not just in efficiency, but in their ability to leverage AI's full potential.

Realignment In How Products Create and Capture Customer Value

Just as cloud computing transformed software pricing from transactional licenses to per seat subscriptions, AI is catalyzing the next evolution in product monetization. This shift goes beyond simple pricing changes—it's changing how products create and capture value. Two key monetization models are already emerging in the AI era:

• Usage-Based Pricing at a New Scale - While usage-based pricing isn't new, AI is redefining what "usage" means. Companies are pricing based on intelligence consumption: queries processed, insights generated, or decisions automated.

• Outcome-Based Monetization - AI enables products to shift from charging for features to charging for results. Rather than paying for access to capabilities, customers pay for verified outcomes: successful customer service resolutions, qualified sales leads generated, or processing time saved.

As monetization models change, they will demand different success metrics. Traditional SaaS metrics like ARR will be supplemented by AI-specific indicators. Product teams must now optimize not just for user engagement, but for the efficiency and effectiveness of their AI systems.

The Second and Third Waves Are Yet To Come

The above is just the first wave of where AI is starting to have an impact. But just as in the shift from on-premise to cloud, there will be ripple effects that we don’t see yet. The second and third order effects stand to be even bigger then the first wave.

The challenge right now is that there is a large gap between AI promise and reality of implementing all this change in product teams. This something I will go deeper on in another blog post.

Redefine, Reinvent, Rebuild, Reforge

I started and named Reforge for moments like these. The word “reforge” is about breaking something down to it’s foundational elements, mixing with some new ingredients, and putting back together to make it stronger, faster, better. That is exactly what product teams, including Reforge, will need to do over the next couple of years. This blog post is a small beginning.

We are building tools for AI native product teams such as Reforge Insights. In addition, continue to build out in-depth expert led courses to help you make the transition like: AI Foundations, AI Strategy, AI Growth (coming soon), and AI Leadership (coming soon).