Tweeted in January that the technical gap is closing fast. Six weeks in, it isn't closing, it's closed.
Claude Code, custom skills, MCPs, a quiet morning and I have what a five-person team had three years ago. I'm running two products this way. The bottleneck is never the code. The bottleneck is that nobody knows you exist, and even if they did they wouldn't trust you yet.
What got cheaper. Shipping a new feature, a day, sometimes an hour. Migrations, refactors, tests, a fraction of what they used to cost. Documentation, marketing pages, internal tools. The thing you used to hire a junior to do.
What didn't get cheaper. One stranger reading one of your posts. One stranger trusting you enough to sign up. One stranger trusting you enough to pay. One referral from a real conversation. The patience to be small in public for six months while it compounds.
LLMs don't help with any of that. They make text production faster, which makes the feed louder, which makes attention scarcer. The net effect on distribution is negative for most founders.
The bar moved. Five years ago "can you build it" filtered out most would-be founders. That filter is gone. The new filter is "can one stranger find a reason to care." That filter rewards different skills. Writing in public consistently without becoming generic. Doing the unscalable thing for fifty people you picked by name. Showing up in comments without selling. Being specific enough to be wrong, not safe enough to be ignored.
The technical founders who are winning now look more like writers than engineers. Not because writing is the only skill that matters. Because writing is the skill that proves you have a point of view, and a point of view is the only thing the algorithm can't generate.
Don't outsource distribution to the model that made distribution harder. Don't generate fifty LinkedIn posts to find one good one. Generate fewer, think more, post the one you'd defend to a hostile audience.
The arbitrage isn't typing faster. It's having something worth typing.
— Simon