For decades, the startup world repeated one mantra: “Ideas are cheap. Execution is everything.”
That was true when execution required engineers, capital, and months of runway. It made sense to worship the builder — the hero who could ship when others could only dream. The conditions of that era made execution the scarce resource, and the framing served us well.
Those conditions are now ending.
The Wrong Mental Model
The common forecast for how AI reshapes business is “solopreneur enterprises”: one person with a fleet of AI agents replacing an entire company. This is directionally true but still thinks inside the wrong frame. It assumes the future still has companies — just smaller ones.
The deeper shift is that AI is removing the barrier between having an idea and shipping it across the entire lifecycle of a business. Coding, design, marketing, sales, support, operations — the full stack of execution that defines software companies is being compressed into natural language and automated reasoning. What used to require a team, a codebase, and a cap table now requires a prompt and an afternoon.
This applies even more to digital businesses, where the product is bits, not atoms. When execution gets this cheap across every function, the old scarcity model inverts. The bottleneck was never ideas. It was the cost of making ideas real. Remove that cost and execution stops being the differentiator. Ideas become the scarcest resources left.
The Defensibility Death Spiral
Here is where the market is going.
First, AI is letting anyone build software using natural language. This is not just about developers becoming more productive — it is enabling even people without technical backgrounds to create complete products. This means, more builders are entering, building more products, competing for the same user attention. The obvious response is a race to the niches — smaller slices, loyal users within extremely differentiated verticals.
But the endgame is sharper than most people expect. Once AI agents are ambient and accessible to every user, the long tail will collapse under its own weight. Why subscribe to someone’s tool when you can simply describe what you need to your own agent and have it generated on the spot, exactly to your specs?
The software company, as a durable entity selling a fixed tool to a stable user base, is approaching obsolescence. Not gradually. At the limit, code has no moat. Software as an industry is trending toward zero marginal cost of creation. The thing companies used to sell — a finished product — is ceasing to be scarce entirely.
This does not mean nothing survives. It means what survives is something else.
What Survives
When execution is commoditised, the creator’s judgment becomes the primary compression function for meaning.
Think about film. Cameras, editing suites, and VFX tools are available off-the-shelf. Anyone can buy them. Production crews, equipment, and capital can be swapped between projects. Yet a “Christopher Nolan film,” a “Studio Ghibli film,” a “Quentin Tarantino film” — these phrases mean something precise. Capital flows to those directors and producers because of their judgment, not the other way around. Money is fungible. Talent is not. The director’s economic value is not in building — it is in deciding what gets built and earning the audience that trusts that decision.
This is the shift coming to digital companies. They will not build the software product, service or tools. They will curate the things that fit their world view. They will describe what should exist. They will opine on how it should be built. But the building will happen locally, where the users live and where their private data resides.
The Edge That Matters
If software itself has no moat, what does?
Distribution.
The asset is the group of people who trust your judgment about what is worth building, worth using, worth paying attention to. It is built on trust. That trust compounds slowly and is impossible to replicate, because it is inseparable from the person whose judgment formed it.
When so many other advantages — technical skill, feature velocity, capital, team size — are being flattened by AI, this is the most durable structural advantage left. It is also the hardest to build, because it cannot be bought, hacked, or delegated.
In the short and medium term, we will keep building software. We will ship products, chase product-market fit, and sell subscriptions. There is no shame in this — it is the reality of the current market, and it is how judgment earns its proof.
But do not confuse the temporary scaffolding for the building. The product is only the evidence of your judgment. The real asset is the audience that forms around that proof.
So, build things that are unmistakably yours. Speak to a point of view that someone could disagree with — and do not try to cater to the majority, as designers and product teams do today. Create work where people sense the hand of a specific mind behind it, even if that hand was amplified by AI.
In the long run, the individuals who endure will not be the ones with the best AI stack or the most polished app. They will be the ones who built the distribution channel no one else can buy: a body of work, a perspective, and a group of people who believe them when they say, “This is worth your time and attention.”
That is the edge that matters.
Discussion