From Feature Factories to Company Factories

We went from teams shipping features endlessly to AI pumping out code autonomously. The next leap — where a single founder builds and runs an entire company with a swarm of agents — isn’t science fiction. It’s the near future. And the people best positioned to seize it might surprise you.

The Three Eras

Era 1: The Factory We Built — and Outgrew

For most of the last decade, the dominant failure mode in software product development had a name: the feature factory. Teams would grind through backlogs, ship tickets, measure velocity, and declare wins — all while users kept churning and the product drifted further from any coherent strategy. Marty Cagan and others spent years warning about it. Most organisations ignored them.

The feature factory was a product of constraints. Building software was expensive and slow, so you optimised for throughput. You hired PMs to manage queues, designers to add polish, engineers to ship. Everyone had a lane. The system made sense given the inputs.

Then the inputs changed.

Era 2: We Are Living in a Code Factory

Since late 2024, something remarkable has been happening quietly inside engineering teams. The arrival of long-horizon agentic coding — tools like Cursor, Claude Code, Factory’s Droids — shifted the fundamental unit of work. Engineers stopped writing most of the code and started reviewing, directing, and orchestrating it instead.

In early 2026, StrongDM’s engineering team publicly described their approach in a post covered by Simon Willison: a “Software Factory” running non-interactive development where specs and scenarios drive agents that write code, run test harnesses, and converge — without human review of the code itself. Their benchmark: “If you haven’t spent at least $1,000 on tokens per human engineer today, your software factory has room for improvement.”

This isn’t just a productivity tool. It’s an architectural shift. By December 2024 — with the second revision of Claude 3.5 — long-horizon agentic coding workflows began to compound correctness rather than error. Gartner projects that 75% of enterprise developers will be using AI tools in their work by 2028. That estimate now looks conservative: we’re already there for any team paying attention.

Key data points:

• 25% of Google’s code is already AI-generated (Sundar Pichai, 2024)
• Solo founders using AI complete tasks 55% faster, with 22% lower capital requirements
• AI coding agents are beginning to displace the $250B annual US software engineering spend

The feature factory isn’t dead. But it’s being automated. The question of what to build remains deeply human. The question of how to build it is increasingly delegated to machines.

Era 3: The Company Factory Is the Destination

If Era 2 is about automating code, Era 3 is about automating the company itself — or rather, reducing the operational overhead of running one to something a single, capable human can manage with a swarm of intelligent agents.

The signals are everywhere. Sam Altman has publicly talked about a betting pool in his CEO group chat for when the first one-person billion-dollar company will appear:

“We’re going to see 10-person companies with billion-dollar valuations pretty soon. In my little group chat with my tech CEO friends, there’s a betting pool for the first year there is a one-person billion-dollar company — which would have been unimaginable without AI.”
— Sam Altman, CEO of OpenAI

Dario Amodei, at Anthropic’s Code with Claude conference, put a 70–80% confidence level on it happening by 2026 — pointing specifically to proprietary trading, developer tools, and automated customer service as the initial proving grounds. Mike Krieger, who built Instagram to a billion-dollar exit with 13 people, said plainly: “I think now you’d be able to do a better job than we did with AI.”

What does a company factory actually look like? It’s a founder — one human, armed with deep product and domain judgment — directing a coordinated network of AI agents acting as functional leads. Not autocomplete. Not chatbots. Agents that plan, execute multi-step workflows, iterate based on outcomes, and hand off to other agents.

The founder remains the Chief Visionary Officer. The AI crew handles technical architecture, full-stack development, content creation, campaign execution, customer onboarding, financial forecasting, and support triage. Operations that once required entire departments are handled by coordinated agents that adapt in real time.

A note on the “solo billionaire” thesis

Is this a stretch? Perhaps marginally — the first wave will likely produce solo unicorns before solo billionaires. But the trajectory is clear. AI unicorns in 2024 reached billion-dollar valuations with around 200 employees on average per CB Insights — down dramatically from prior cycles. Over 30% of new unicorns in 2024 used AI as a core component. That headcount number is compressing fast. The solo billion is a question of when, not if.

Historical precedent matters here too. Instagram was valued at $1 billion in 2012 with 13 employees. WhatsApp sold to Facebook for $19 billion with just 55. Mojang (Minecraft) sold to Microsoft for $2.5 billion with about 40 people. The direction of travel is unambiguous.

The Skills That Unlock This Future

Here’s what often gets missed in this conversation: the Company Factory era doesn’t reward generalists who dabble. It rewards people who have genuine depth in multiple domains — what some call “T-shaped” but is really more of a “π-shaped” profile, with two deep verticals instead of one.

The company factory demands that a single person can cover what traditionally required a VP of Engineering, a Head of Product, a Growth Lead, and a Marketing Director — not by doing everything manually, but by being credible enough in each domain to direct AI agents effectively and catch when they’re wrong.

Engineering Practice

You don’t need to be a 10x engineer. But you need to understand systems. You need to read code well enough to catch agent errors, design architectures, and make tradeoffs. In the code factory era, engineering fluency is table stakes for anyone who wants to build. The new literacy isn’t just prompt engineering — it’s systems thinking expressed through code, even if agents do the typing.

Product Management

This is the most underrated skill in the company factory stack. PM is fundamentally about judgment under uncertainty — identifying the right problem, de-risking the solution, and closing the loop between user behaviour and product direction. AI can generate features. It cannot tell you which features matter. It cannot hold the user’s Jobs to Be Done in tension with the business model. That remains irreducibly human.

Marty Cagan at SVPG has been making this point since before agents existed: the PM role was always about being a product creator, not a backlog manager. His framing: high-agency PMs who understand value and viability are “ideally suited for the AI-powered future — and you can already see the market rewarding these people with rapidly increasing salaries.”

Growth & Marketing

Distribution is leverage. An exceptional product with no distribution is a dead product. The company factory founder needs enough marketing and GTM fluency to direct AI agents building content pipelines, running ad campaigns, mapping customer journeys, and crafting positioning. You don’t need to write every blog post. You need to know what makes one land.

Domain & Customer Depth

AI is a force multiplier on insight, not a substitute for it. The founders who will build company factories aren’t generalists — they’re specialists with reach. They understand a specific customer’s pain well enough to direct every function of the company toward solving it. JTBD methodology, customer interviews, outcome-based thinking: these are durable and AI can’t replace them.

High Agency

Vercel’s internal motto is “you can just ship things.” That disposition — bias for action over process, willingness to be the constraint rather than wait for one to be removed — is the connective tissue that makes the other skills compound. Without it, even a π-shaped founder becomes a spectator.

Why Product Managers with Engineering Practice Are the X-Factor

There’s a specific combination that maps almost perfectly to what the company factory requires: a product manager who can code. Not a “technical PM” in the watered-down sense of someone who attended one engineering all-hands. Someone who has shipped products, debugged production systems, understands the stack, and can have credible conversations with — and give clear direction to — AI coding agents.

Reforge put it plainly: AI is giving PMs the opportunity to “become a really good designer — or get a little more dangerous with code.” The Head of Design at Cursor went further, writing that what companies want now are “T-shaped builders — people with deep expertise in one area and enough breadth to contribute to many others.”

The compound advantage is structural. A PM with engineering practice can:

• Close the loop between insight and implementation without a handoff
• Prototype to validate before committing engineering resources
• Direct coding agents precisely because they understand what the agent is actually doing
• Review agent output without delegating that judgment to someone else

In a company factory, every handoff is a point of failure. The PM-engineer hybrid eliminates the most critical one.

The hiring signals confirm this. As one product manager wrote recently: “AI companies are explicit about this — the only credential that matters to AI-native teams is what you’ve actually built and what tangible ideas you have to improve their product.” Josh Woodward, VP of Gemini at Google, framed the new interview filter simply: “What are you building in your spare time? People who like to tinker and build — they express themselves through prototypes, not docs.”

This is not a coincidence. It’s a description of exactly the skill profile that the company factory demands.

The Honest Caveats

None of this means every solo founder with AI access becomes a billionaire. The company factory model has real constraints today. AI agents make inhuman mistakes. They hallucinate architectural decisions, introduce subtle bugs, drift without tight context management.

Tom Coshow, senior director at Gartner, warned in 2024: “With LLM-based AI agents, the situation is such that we have to give them very simple decisions to get reliable answers. We are nowhere near the point where we can just throw a lot of data at an AI agent and trust its decision.” That was 2024. The trajectory suggests it won’t always be true — but it’s true now, which means the value of the human orchestrator is highest right at this moment.

The first solo companies to reach unicorn valuations will likely be in proprietary trading, developer tooling, and automated services — not because those are the only domains, but because they have the clearest feedback loops, the most automatable operations, and the least need for physical-world coordination. The company factory model generalises from there.

What This Means If You’re Building Now

The transition from code factory to company factory is not a future event to wait for. It’s a direction to orient toward now.

The practical implication: if you’re a product person, the highest-leverage investment you can make is engineering depth. If you’re an engineer, it’s product judgment and customer proximity. If you’re a founder, it’s systematically replacing human operational overhead with agent-based systems — not to save money, but to compress the feedback loops between insight and impact to near zero.

The feature factory burned out teams by optimising for throughput over outcomes. The code factory risks the same trap at a higher level of abstraction — shipping more code faster without asking whether the code solves anything real. The company factory, done right, closes that loop permanently: one person with deep judgment, broad leverage, and a crew of agents executing with precision.

The first solo billionaire isn’t a prediction. It’s a direction of travel. And the map has been visible for long enough that there’s no excuse not to be moving.