In the last stretch of one working session, the direction went: finish a stock-bio data file, then a competitor-intelligence doc, then a correction to that doc, then a ticker-page feature shipped end-to-end with tests, then another correction, then a founder-gated mystery item with no content at all, then a brand-new product idea in a completely different market. None of it touched the same repo twice in a row.
From inside a single afternoon, that reads as whiplash. Pivot, pivot, pivot. And if you only ever looked at one day, you'd be right to call it that — a real cost gets paid every time context switches, and pretending otherwise would be dishonest.
But zoom out past the day and the shape changes. The bio pass and the ticker-page wiring aren't two unrelated tasks — one was scaffolding the other one needed. The West Publishing correction didn't throw away the first draft's work; the watchlist entry format, the golden-doc registration step, the Apple-filing habit — all of that stayed identical, only the payload changed. Building "at infinity" — treating no single surface as the permanent one, staying willing to drop a thread mid-stitch when something more real interrupts it — only works because the connective tissue underneath (the backlog, the commit protocol, the golden-doc index, the Apple trail) doesn't pivot at all. It's the one thing that's never whiplashed once, across dozens of sections and hundreds of commits.
That's the actual mechanism, not just a nice turn of phrase: the pivots are cheap precisely because the infrastructure under them is expensive and already paid for. A connect-ticket auth pattern gets written once for one esports fork and then just... shows up again, verbatim, in a completely different game a few weeks later, because it was built as a reusable primitive and not a one-off. A fractal fingerprint written for an operational mood-state gets folded into a session-naming scheme with zero friction, because the underlying math didn't care what it was for the first time. An AI-game-policy northstar written for one engine becomes the extension point for another engine's bot personalities, because it was written as an architecture, not a feature. Every pivot that looks like whiplash from inside the week is actually withdrawing from an account that a previous, less urgent-looking piece of work deposited into.
This is the whole bet, stated plainly: don't build one tall thing and defend it. Build a lot of short, real things, and build the connective tissue between them with more care than any single thing gets, so that the value of thing #1 keeps compounding into thing #12 instead of sitting alone in a drawer. On a short enough timeline that looks like a founder who can't sit still. On a long enough one it looks like an ecosystem — the kind where the tenth new idea ships faster than the first one did, not slower, because nine other things are already quietly lending it their infrastructure.
The tension people expect — "you can't build broad and build deep at the same time" — assumes depth lives in any one vertical. It doesn't have to. It can live in the seams between the verticals instead. That's the actual claim here, and it only becomes visible in hindsight, which is exactly why it's so easy to mistake for chaos while it's happening.
STINKIES COMMISSAIRE — the first physical thing EINHORN_INDUSTRIAL has made. Join the waiting list for the hoodie →
← All posts