Why the mean era of software is ending
In late 2024, a burned-out founder in Vienna started building tools for himself.
A tool to control his terminal from a browser. A bot to answer his WhatsApp. Scripts to automate his house. Dozens of apps, each one built for an audience of exactly one. By every rule of software economics taught for twenty years, this was a waste. No market. No cohort. No TAM slide. Just a rich man's hobby.
Eighteen months later, those toys had connected into OpenClaw, one of the fastest-growing projects in GitHub history.
The rules weren't broken. The rules had changed. Nobody sent the memo.
The Audience Tax
For most of software's history, building has been brutally expensive. Engineers, designers, infrastructure, quarters of runway: every product carried a heavy fixed cost before a single user touched it.
Expensive things need to justify themselves. Software justified itself the only way it could: with audience size. If it cost this much to build, it had to matter to this many people. Every product had to have a market.
Call it the Audience Tax. Every idea had to survive one question before a single line of code was written: but will enough people want this?
Here's the problem: most problems faced by individual humans don't survive that question. Your actual workflow. Your family's actual schedule. The way you think about your money, your notes, your mornings.
These are real problems that failed the filter because the audience was one.
So we never built for a user. We built for a market. That single choice shaped everything downstream, including our most sacred acronym: the MVP.
The Mean Viable Product
To pay the tax, you couldn't build for a person. You had to build for the statistical center of a cohort, the average of a thousand imagined users.
Not the minimum viable product. The mean viable product. Software shaped for nobody in particular.
Because it fits no one exactly, we built an entire vocabulary to manage the mismatch:
- Onboarding: teaching people to want the mean.
- Feature requests: people begging for their edge back.
- Power users: the ones who bent the mean by force, with plugins, hacks, and duct tape.
The machine that produced all this had a name too: the Funnel. You know it well. Start wide. Pick a large cohort because the tax demands one. Guess at its center. Ship the mean. Then spend years trimming and growing: cut the features the middle ignores, add the ones it begs for, and sculpt your way toward that mythical state of grace, product-market fit.
Wide, then narrow. Cohort, then fit. Mean, then meaning.
None of this was wrong. It was the only rational response to the economics. When building costs quarters and millions, you cannot afford to be wrong about the audience, so you aim for the fattest part of the distribution and pray.
But rational responses to a constraint die with the constraint.
The Repeal
You know what happened next. It's happening now.
LLMs and agentic coding collapsed the cost of building by orders of magnitude. What took a funded team two quarters now takes a builder a weekend. The majority of people building software today wouldn't have called themselves developers three years ago.
The ROI math now works at n=1.
A few people saw this coming. Robin Sloan, back in 2020, argued an app could be a home-cooked meal, made for your family, where "why don't you scale it?" is as absurd as asking a home cook why they don't open a restaurant. Maggie Appleton predicted a golden age of barefoot developers, embedded in their communities, building basic software care for the people around them.
They were right, and earlier than everyone. But notice the frame: personal software as an escape from scale. Home cooking as the alternative to the restaurant.
I think the more interesting claim is stranger than that. Personal software isn't the escape from scale. It's the new starting point of it.
The Inverted Funnel
The old funnel ran wide to narrow. The new one runs narrow to wide.
You start at n=1. Software so tailored it's almost embarrassing: your workflow, your edge cases, your taste baked into every default. No settings page, because you are the settings.
I've been living this. My entire game right now reduces to one metric: how short is the distance from idea to usable toy? Not shippable. Not scalable. Crafted for me, by me, today.
Which means I don't run a product org anymore, even in my head. I run a home kitchen. A rotation of small recipes, cooked for home consumption. Like any home kitchen, the real leverage is in the hacks that travel. A trick invented for one dish gets remixed into three others. Recipes compound. The tenth toy costs a fraction of the first.
My kitchen has three rules.
- Whatever is cooked gets eaten: if I won't actually use it, I don't build it.
- It has to be fun: joy isn't a bonus here; it's the fuel that keeps you building long enough for the connections to appear.
- Be stingy with time: the entire economics rests on being wrong staying cheap. Protect the cheapness.
These sound like hobby rules. They're not. They're the discipline of the singular path, the difference between a drawer of abandoned scripts and a portfolio of living nodes.
Something changes when you build at n=1, and it's not just the cost. It's the feedback loop.
In the old funnel, learning was slow and statistical. Cohort analytics, A/B tests, quarterly reviews: inference about a mean you could never actually meet. In the new funnel, the user is you, or someone you can text. The feedback loop collapses from quarters to hours. You don't infer the problem. You have the problem.
Experimentation becomes free. A niche user story is no longer a roadmap gamble that must fight for a sprint. It's a Tuesday afternoon. Build it, live with it, keep it or kill it. The cost of being wrong has dropped so far that being wrong became a feature.
The old funnel sculpted a mean and hoped it resonated. The inverted funnel starts with resonance, deeply right for one, and lets scale be discovered.
Most of what you build for yourself will never meet a larger audience, and some of it will quietly rot: unpatched, unmaintained, insecure in ways that would horrify a production engineer. The skeptics are right that personal software carries a day-two problem, and the studies showing AI-assisted builders overestimate their own output are worth taking seriously.
But here's what changed: dead personal software used to mean dead capital. Now it means a Tuesday afternoon.
The Queen Bee
Go back to Vienna.
The dozens of personal tools weren't a product. They were nodes, each one true for one user, each one absurd by the old math. They could have stayed that way forever: a drawer full of toys. Nodes don't self-assemble.
What turned them into OpenClaw was one more invention, by itself the least impressive of the lot. A harness. A thin layer that let the tools talk to one another, and then to everyone else's. Not a breakthrough product. Barely a product at all.
It was the queen bee.
A hive isn't ten thousand bees. It's one organism. What fuses it into one isn't the strongest bee or the fastest. It's the one every other bee organizes around. Every singular path that ends in scale has this moment: the unglamorous connector that compounds a pile of tools into something with a life of its own. You can't plan it. You can't know which connective layer matters until the nodes exist to be connected. But you can be ready for it. The builders who reuse their hacks across dishes are the ones who recognize the queen bee when she arrives.
Nobody in Vienna planned an organism. Nobody ever does.
The Singular Viable Product
The old discipline asked: how many people want this?
The new discipline asks: how true is this for one person?
That's the whole inversion. Truth at n=1 is the new seed of scale. The mean viable product had a good run. It was the right answer to an economics that no longer exists. What replaces it isn't a smaller MVP or a leaner funnel. It's a different object entirely: the singular viable product. Built for one. True for one. And sometimes, unpredictably, organically, the first node of something enormous.
Stop building for the mean.
Build for someone.
Start with you.