The Open-Soruce Big Bang – Random Thoughts on Leadership & Technology

click here to subscribe for information on upcoming articles before they are released and periodical summaries of the most read articles.

The Open-Soruce Big Bang

18 Jul, 2026

On why most of your hard problems were already solved by a fish, a mammal, and soon, an open-source solution

The fish and the mammal who took the same exam<br>Put a dolphin next to a shark. Same torpedo body, same dorsal fin, same general business of being the thing other animals do not want to be downstream of. You would assume close cousins. You would be off by about 300 million years and an entire mountain of priorities. The shark is a fish. The dolphin is a mammal that got bored of land, walked back into the ocean, and re-derived the shark from first principles, because the ocean only accepts a narrow range of answers.

This is convergent evolution, and the lesson is a little rude - the environment sets the problem, and the problem allows only so many good solutions. Build something fast in deep water and you will end up looking like a shark whether your ancestors were fish or college-educated land mammals. Nature does not award points for originality. It awards survival, and survival has a style in water.

Evolution, importantly, has no patent office. The shark spent roughly 400 million years on research and development. The dolphin showed up a few dozen million years ago and shipped essentially the same product with a blowhole and a personality. Nobody got sued. The good design was just lying there, in the water, free for anyone willing to die enough times to find it.

The cat that decided to be a dog<br>Now run the tape the other direction. The hyena looks like a dog, runs like a dog, and shows up at the same buffet as the dog. It is not a dog. The hyena sits on the cat side of the family tree, right next to the lion. The lion and the hyena are relatives who clearly could not stand each other at the reunion, took completely different jobs, and grew into completely different shapes to do them. One became the brooding apex influencer who sleeps twenty hours a day and gets all the press. The other became the relentless, bone-crushing operations team that actually clears the plate.

Same family, opposite niches, opposite forms. And in the dolphin case - total strangers, same niche, same form. Either way, the niche wins. Ancestry is merely a suggestion. The problem is the boss.

Business is just convergent evolution with worse smells<br>Now look at what humans build, and notice that we are nowhere near as clever as our pitch decks insist.

Every marketplace independently re-derives ratings, reviews, and a trust-and-safety team that is permanently sad. Every subscription product re-derives the three-tier pricing page with the middle column highlighted, because someone read the same behavioral economics paper you did. Every social app re-derives the feed, the like, the notification dopamine drip, and eventually the apology tour. Every growing company re-derives the org chart, then the reorg, then the second org chart.

We call these things innovations. They are dorsal fins. They are the shape the water forces (in our case socio-technical forces, societal factors and mist if all human behavior) onto anything trying to move through it. The startup graveyard is full of founders who believed their checkout flow was a brand-new genus, when it was in fact the four-hundredth independent evolution of the same button.

"Our proprietary algorithm", in most cases, is a phrase that means "an if-statement we have become emotionally attached to". The real moat is usually just the part of the system the team is too embarrassed to open source, not the part that was ever actually hard.

None of this is an insult. It is the entire point. If problems have a small number of good solutions, then everyone who is any good will keep arriving at that same small number of good solutions. The convergence is evidence the solution is correct. It is not evidence that everyone is a thief.

Enter the convergence engine<br>Here is where it gets funny, and slightly threatening, depending on your equity situation.

AI is, by construction, a machine for the common solution. A large model is trained on the accumulated output of nearly everyone who has ever solved a problem and then written about it, which makes it exquisitely good at exactly the part of the world that has already converged. Ask it for the standard answer and it hands you the standard answer at a speed that makes your senior engineer feel things.

This gets pitched as a weakness - it is mid, it is average, it regresses to the mean. Sure. But the mean is precisely where all the solved problems live. Being a fluent native speaker of the solved-problem space is not a bug when the task in front of you is, statistically, a solved problem wearing a fake mustache.

Agentic engineering then closes the loop. The model no longer just describes the...