The Great Distribution

TL;DR Overview

The core components of this market thesis are as follows:

1. Talent, money, and time were historical moats for large organizations and venture-backed startups. Those constraints also made software engineering more defensible because building good software required access to scarce technical talent, capital, and long execution cycles. That is changing.
2. Maintaining alignment between engineering and business is key to growth, but that alignment often diminishes as teams scale. Addressing this gap is crucial for sustained success in this new market.
3. Modern AI capabilities, despite meaningful roadblocks to production adoption, will reduce the talent, money, and time needed to build a successful company by an order of magnitude in many categories.
4. Product and GTM teams will play an even more important role as catalysts in translating evolving market demands into actionable, competitive strategies.
5. Data and distribution are now your only superweapons.
6. This creates a hyper-competitive market I'm calling The Great Distribution: a period where the largest technology platforms continue positioning themselves at the top of the value chain, while many non-core workflows, products, and markets become increasingly contestable by AI-native startups.
7. Alongside this distribution is an opportunity to create wealth in markets that historically did not have enough talent, money, or time to solve the problems they were riddled with. The most optimistic version of this future is not simply job displacement, but a massive expansion in new company formation, new categories of work, and new entrepreneurs globally.

Talent, Money, and Time

The zero-interest-rate market was a catalyst for big tech's growth and for many key innovations coming to market. Companies were able to raise nonstandard rounds of funding at favorable terms, spend significant time building products with long development cycles, and operate at a loss for extended periods. That allowed several meaningful innovations to enter the world at consumer-friendly prices while organizations that burned a lot of money on talent and time could pursue profitability later.

The biggest laggard to sustainable, on-target revenue growth has historically been building enduring customer value at the speed the market requires. That usually depends on both a strong technical organization and a strong customer-centric operating culture.

Excellent founders capture customer needs, go back to the drawing board, and figure out how to meet those needs. In the past, this was gated by talent, money, and time. With the right people and funding, you could promise features to customers and deliver a fuller product over time. The main pressure was competition.

In today's market, investors have to take on more risk to meet their targets, and that requires higher confidence in a startup's progress, GTM, product velocity, and market execution. The new norm is more disciplined funding, more stringent early rounds, and more pressure to show tangible growth. Startups must demonstrate the potential for higher returns relative to lower-risk investments.

While these market constraints create tougher conditions for founders and investors, there is one variable that de-risks a meaningful amount of this pressure: access to artificial intelligence.

Our New Baseline with AI

Even basic uses of AI (code generation, copilots, research assistance, business development support, prototyping, internal tooling) can materially compress product development cycles.

If AI creates a 2x or 3x efficiency gain in a meaningful subset of software workflows, that does not simply mean "engineers write code faster." It changes the financial structure of a company. It changes the number of people needed to get to product-market fit. It changes the threshold for who can build. It changes how much capital a company needs before it can test a serious product in market.

By distributing access to building products to less technical teams, the variables needed to de-risk a startup begin to change.

We are in a new market. We may not see the zero-interest-rate era return in the same way, and in many software categories, we may not need it. Companies designed before the AI-native era will need to re-architect how they work, how they build, and how they stay close to customers.

A small AI-native team founded in 2023 or beyond can now go after large or niche market solutions in a fraction of the time. These organizations will be structured differently than traditional companies built around large engineering teams and long roadmap cycles. They can deliver faster. They can test more. They can stay closer to customer needs in a way that larger, more process-heavy organizations may struggle to match unless they change how they operate.

The R&D curve in AI will continue making this more possible. Multimodal models, memory, recall, agents, workflow automation, and continued human feedback will keep shifting the baseline. Many software development tasks will become more homogeneous. Software engineers will increasingly operate as copilots to AI for a growing percentage of product work. Others will focus on complex engineering challenges where deep technical expertise remains the differentiator.

Global mobility may also increase as developing and underserved countries get a better shot at attracting and retaining technical talent. If building software requires less capital and fewer specialized local networks, more people can build from more places.

That alone changes everything.

Longer-Term Market Outlook

With a massive increase in access to becoming founders, building products, and capturing value, there is a transformational growth opportunity ahead.

We may see global GDP growth in categories and regions that were historically constrained by access to capital, talent, and technical infrastructure. Today's working class, even in underserved countries, increasingly grew up with smartphones, internet access, and exposure to global markets. The localization of AI to regional market challenges can bring powerful technology within reach of the masses.

These productivity gains could create more wealth than previously imaginable. In many categories, you may only need a laptop, market insight, distribution, and access to enough compute to approach an industry, build better workflows, ship faster, and increase profits.

Access to compute will be a major challenge. Hardware efficiency and optimization will become an even more important engineering focus as demand for data centers, GPUs, energy, and networking infrastructure grows. Countries constrained by compute, energy, capital, or cumbersome compliance requirements may eventually hit a wall and need to decide how aggressively they want to participate in this opportunity.

This makes sovereignty an even greater challenge and an even more important capability for countries to pursue. There's a whole writeup incoming on sovereignty and owning your future.

Roadblocks Ahead of Production Adoption

The future we're charting with AI comes with significant roadblocks.

Although nearly every organization today is describing its products as AI-powered, the latest large-scale interpretation of value comes from the adoption of language models, agents, and AI-enabled workflows. These systems still have issues that prevent broad real-world adoption in many business contexts.

Language models are powerful at extracting patterns from corpus of data and translating them into useful knowledge in natural language. But we still have not fully reached the inflection point where they are trustworthy enough for general production adoption across every business-critical workflow.

LLMs are inherently non-deterministic. For the last 70 years or so, we built computing systems on the premise that we could define an expected output and then use enough talent, money, and time to engineer toward that output. Those outputs were determined by predictions of customer needs or direct customer feedback. Today, AI reduces the talent, money, and time needed to produce those same or greater outputs, but it also introduces new uncertainty.

The key factors preventing wider production adoption include:

• Hallucinations. Production applications often have low tolerance for error. The margin of error may be greater, unspecified, or difficult to evaluate, and those errors can create uncertain risk.
• Evaluations. To mitigate errors, we need evaluations that assess model quality based on alignment to performance datasets. This is still riddled with challenges: using AI to do evaluations, human bias in assessing quality, construction of performance datasets, and the lack of broader inspection criteria beyond quality alone. A true enterprise-grade inspection needs to consider quality, security, privacy, performance, reliability, and cost. Call it a six-point inspection.
• Reliability. Enterprise and customer-facing applications often come with assurances like SLAs, SLOs, and SLIs. Promoting a model into a critical user-facing workflow without predictable reliability is a non-starter for many production applications.
• Cost. Turbo models, caching, distillation, routing, and mixture-of-experts architectures are bringing costs down, but there is still a long way to go before inference is economically feasible across every organization and workflow. AI costs can quickly start to look like cloud spend. Large organizations may not always be structurally prepared to capture the same margin profile as AI-native companies.
• Data strategy. Organizations are already reconsidering how they protect their data: blocking crawlers, revisiting what they publish, constraining where internal knowledge lives, and controlling how employees use AI tools. This may lead to a world where frontier models face constraints around fresh, proprietary, or high-quality knowledge. I expect smaller and more specialized models to rise in importance, including industry-specific models built from differentiated R&D and proprietary data. (Think of a Mayo Clinic offering a specialized frontier model to the market.)

On top of all of this, we are running into ethical, safety, IP, and copyright discussions. Training on the largest possible corpus of data and assigning weights to different datasets creates a unique dilemma for society. Is our internet-scale data already predisposed to a level of bias that needs to be resolved before we compound it with additional biases in training data and model behavior?

If we massively distribute these compounded biases through AI, will that reduce the breadth and depth of unique thought?

The possibility of plateauing the expansion of our own minds is worth considering.

We will need to strongly consider how to unbias data, models, systems, and even the authors of content, especially as we move toward increasingly general systems.

I believe there are and will be enough researchers and engineers to solve many of the roadblocks to AI adoption. But in parallel, we need modern thinkers to step outside the existing scope of data and operate from the unknown, underserved, and underreported parts of humanity.

That is how we maximize the utility of our genius human species.

What Does This Mean for Big Tech and Enterprise?

We are already seeing the beginning of a restructuring period for large enterprises and organizations not yet optimized for an AI-centric world.

Many organizations have gone through layoffs, restructurings, reprioritizations, and operating-model changes. Some roles will not come back in the same form. Some engineers and operators will find themselves searching for employment in different types of companies. Pay scales may compress for some roles and expand dramatically for others, particularly where the work is closer to frontier research, infrastructure, AI systems, data, and mission-critical product execution.

Some people will reskill into hardware engineering, robotics, bioengineering, climate, defense, manufacturing, and other emerging markets. Others will start companies. Others will pursue more stable jobs, move to lower-cost cities, or re-evaluate what they want work to be.

Enterprise is also reprioritizing investment toward core competencies. This creates less talent, money, and time for smaller or less strategic parts of the business. We have already seen major technology companies restructure teams, reduce investments in some areas, and shift how they fund experimental work.

This opens up a market for laid-off employees, operators, and founders to build AI-native companies around workflows or customer problems that are no longer top priorities for incumbents. These companies can be smaller, faster, and closer to the customer.

That does not mean incumbents are doomed. In fact, the largest technology companies are strategically positioning themselves to capture enormous value from AI. They have distribution, data, cloud infrastructure, research labs, capital, partnerships, and access to compute. Many will move up the value chain and become even more powerful.

But outside the absolute top of the value chain, many enterprise workflows will become more contestable.

The opportunity for founders is to serve customers in places where the incumbent cannot move fast enough, personalize deeply enough, or justify enough internal priority.

There is no free lunch, but there is a real opening.

Enterprises will mitigate these risks in the short term by optimizing income statements, restructuring teams, making strategic M&A deals, investing in AI capabilities, and using compliance, procurement, security, and vendor trust as advantages.

Some of those advantages are real. Some are temporary. Some will become less defensible as AI-native companies mature.

Consumers and business users increasingly want better, faster, more personalized software experiences. Large enterprises can absolutely deliver those experiences, but only if they are willing to reform how they build, how they buy, how they use data, and how close they stay to customers.

At this rate, the market may increasingly segment itself: you go to the largest platforms for infrastructure, compute, models, research, and scaled trust; you go to AI-native startups for the sharpest product experience in a specific workflow.

Simple thoughts for enterprise:

• If you thought you were customer-centric, multiply that by 100. That is where your incoming competitors will be.
• Be good to your customers. Extremely good.
• If a product line is not strategic enough for you to serve with excellence, consider whether it should become a partnership, acquisition target, spinout, or open market opportunity.
• If you believe compliance, procurement complexity, or vendor lock-in is your only defensibility, know that this advantage may be temporary.
• Software defensibility is shrinking in many categories.
• You will need new revenue channels, new areas of distribution, and new ways to get closer to the top of the value chain. For example, some enterprises may commercialize industry-specific models or AI systems built from proprietary R&D.
• If you do not take a serious leap toward realizing the value of AI, you may never have enough time to be truly excellent for your customers.

Consumer Trends and the Third Wave Coffee Effect

One of the biggest blind spots among industry pundits is a lack of foresight around the next generation of consumers.

Gen Z and Gen Alpha are digital-native in a way no prior generation was. They grew up with devices, software, global content, instant search, social media, and algorithmic personalization. They have a natural affinity for data analysis, selective and often ephemeral interests, and most importantly: they are not patient.

They did not have to wait in line at the DMV in the same way. They did not need a calling card to speak to a friend overseas. They were not constrained by where they had to shop or what local societal norms they had to subscribe to. Pair this with exposure through globalization and social media, and it becomes clear they are less likely to accept old, slow, archaic structures without questioning them.

They want increased personalization, faster solutions, better products, and they want them now.

As mobility and connectivity continue to expand, consumers are increasingly selective and discerning in their purchasing and adoption behaviors.

This shift is similar to what I call the third wave coffee effect.

At one point, Folgers and Maxwell House were among the major brands delivering consistent coffee at scale. They had global supply chains, large organizations, and distribution. In essence, they were like traditional large monolithic software companies: a working product, broad reach, but limited iteration against a developing market.

Starbucks introduced a more experiential version of coffee to the world and revolutionized the industry. They demonstrated that by building a better product, capturing nuanced customer needs, and creating stronger feedback loops, you could increase the depth of a market. You could expand TAM, increase customer utility, and make the category feel different.

In software terms, Starbucks felt like CI/CD reaching the world.

Today, Gen Z and beyond spend more time exploring third wave coffee shops that are more personalized to local consumers and more receptive to local market needs. They discover these products through influencer marketing, social search, and short-form content. They find drinks, spaces, and experiences that align more precisely with their taste. They like certain shops for the vibes.

These shops are not constrained by the same expectations as public shareholder markets. They can serve a smaller customer base with more specificity.

In that sense, third wave coffee shops are a small representation of how AI-native companies can capture value in a distributed market. They are hyper-personalized experiences, highly receptive to customer needs, and often discovered through modern distribution channels.

The acquisition of brands like Blue Bottle by larger incumbents also points to one likely enterprise response: M&A as a way to absorb differentiated customer experience and maintain market position.

TL;DR: the vibes are here to stay.

In Comes The Great Distribution

The Great Distribution will be categorized by the following:

• Market conditions that increase pressure on large enterprises to focus on core competencies and reduce investment in non-core focus areas.
• A dramatic reduction in capital and operational costs for startups.
• A faster path to product-market fit in many software categories.
• Restructurings caused by product deprioritization, automation, margin pressure, or market expectations.
• An evolving labor market with diminishing loyalty to large corporations.
• A consumer base increasingly drawn to superior and personalized products.
• Increased accessibility to building products, including for non-technical and semi-technical founders.
• A shift from over-indexing on the CTO as the center of company formation toward a broader balance between technical execution, CEO-level judgment, product taste, and GTM capability.
• The rise of AI-native companies capturing market share in workflows, products, and customer segments that are underserved by larger enterprises.

In this new business climate, enterprises may struggle to compete across too many arenas at once. Smaller, more agile AI-native companies can focus on specific workflows and customer pain points with greater intensity.

There may be ten founding teams for every one problem, where in the past there may have only been two.

Ultimately, this deflationary period may not necessarily mean broad market deflation. It may be a distribution of business activity across many more startups, leading to faster cycles of product-market fit and more innovation globally.

AI has historically been presented through a doom-and-gloom lens. The techno-optimist view is different: society can transform in ways that reduce dependence on a single country, company, or labor pool while increasing the strength of civilization through a distributed yet interconnected, high-performing society.

AI-Native Companies and a New Org Structure

For startups building in today's climate, you may realize very quickly that the talent, money, and time needed to build a business can be meaningfully lower than what was required before accessible AI.

You can bootstrap with more confidence now.

In some categories, $250K may be enough to get your product in the hands of your first customers. You may also be competing against three times as many competitors. If your technology is roughly the same, your differentiator will be your market presence, execution, distribution, and customer intimacy.

Now is the time to deeply consider the value of effective GTM capability. That may be your moat against other AI-native companies.

Your path to product-market fit should be faster than before. You can use margin to build competitive pricing, offer incentives to drive acquisition, add new product offerings, invest in R&D, or improve your talent funnel.

All things said: technology will not be your differentiator in software unless you are going after a truly difficult problem space. Execution, focus, and proximity to customers matter more than ever.

Frontier labs and model providers will continue looking for ways to improve models, acquire high-quality data, and move deeper into enterprise workflows. At the same time, enterprise-grade controls, contractual protections, and privacy commitments will become increasingly important. The real risk for companies may not be formal enterprise agreements, but unmanaged AI usage, shadow workflows, unclear data policies, and poor governance.

Organizations need to think seriously about their data strategy. As software gets commoditized, differentiated data and proprietary workflows become foundational IP. If your knowledge collapses into the model layer, your moat may collapse with it.

As mentioned earlier, software engineering was one of the greatest enablers of building good products. It gave FAANG and venture-backed startups a major edge in the market. While AI-driven efficiency and productivity gains are a major aspect of the new baseline, the bigger shift is that product quality, velocity, and rework can all improve at once.

This means Product, GTM, and business-oriented roles become more influential in winning the hearts of customers as faster iteration cycles become possible.

In comes the rise of the Technical Product Manager: a super PM who can orchestrate customer and business requirements while deeply integrating with software teams through AI.

These people may become rockstar founders, intrapreneurs, and early-stage hires. They may become general managers one day. They can speak to customers and the business while giving engineers a better starting point to drive alignment, velocity, and quality.

These people already exist today. They will be in the spotlight tomorrow.

AI Paves the Way for More Jobs and a Better Future

Doctors could have developed complex vaccines and identified undiagnosed illnesses but the market timing wasn't there. We could've landed on Mars but the market timing wasn't there. We could've solved the water/energy supply crisis but the market timing wasn't there. We would have invested in all underdeveloped areas of the planet to eliminate hunger, drive education, and decrease inequality but the market timing wasn't there. The focus on what's going away with AI without presenting the perspective of what's at stake for humanity is a perspective that lacks foresight.

The IMF recently predicted 40% of jobs will be at risk due to AI, deepening inequality. While this can certainly hold true with the power of job-scale automation, we can unlock another 140% of jobs in problem spaces we've never historically had the bandwidth to align talent, money, and time towards. We believe on the other side of this 40% is the high probability of decreasing inequality, of driving global sustainability, of helping everyone pursue their hierarchy of needs and achieve their fullest form, of becoming a multi-planetary species rich in culture. The only way that will happen is if we democratize the ability for those same humans, with inherent differences, to bring their ideas into the world, create more jobs, and solve our next generation of problems. This new paradigm of talent, money, and time will change the world for the better.

— Iman