The Cambrian explosion occurred roughly 538 million years ago. It marks a period where life went from simple single-celled organisms to complex multi-cellular life and began to differentiate into nearly every major animal group. Changes in the environment led to this rapid burst of new species creation.
Evolution occurs when genetic diversity meets selection pressure. If there are 10 variations of a given life form and they are exposed to evolutionary pressure (say disease or environmental challenge), the fittest will survive this pressure and propagate their genes. In this process, nearly all mutations are harmful or neutral. Evolution is highly random and isn’t a staircase that only goes up. What leads to new, successful types of life are the rare mutations that improve fit with the environment.
We’re seeing a Cambrian explosion in software today. With improvements in LLM-based AI and AI harnesses, the amount of new software being created has increased massively. In the software analogy, selection pressure means creating enough value for users and the creator that they continue to use it. Admittedly, not all of what’s being made is good. Much of it is not. Low-quality AI slop is real, and we’ll have to deal with the noise that comes with the volume. Yet in the long run, high-value software, both new products and adaptable incumbents, will be what persists.
I have started a lot of side projects over the years, but most of them end up not seeing the light of day: I get busy, bored, or stuck on some engineering challenge. I’m not alone in this; it takes a great deal of skill and grit to finish and polish software for the world. Previously this difficulty filtered for people who could finish projects, not necessarily for the best ideas. AI lowers the coding-and-shipping-competence filter in favor of a stronger market filter.
Now people can have an idea, have AI prototype it, and ship it in a day. Sure, a day only gets you something simple, but it’s a dramatic shift from before. I’ve recently shipped a few side projects, one that’s only intended for an audience of about 7 people — a book group management app to rank and organize the books we read — and a second for managing D&D campaigns as a DM. Both have been valuable to me, and I hope to their users. I know other developers writing small apps to make subtle improvements in their lives, like managing a calendar for their family or keeping track of their fitness goals. Lots of creative people who didn’t have the time or skills before are now creating things.
Similar to what happened following the Cambrian explosion period, much of the new diversity won’t survive as it is culled in favor of the more successful species. The point isn’t to maintain diversity forever, it’s to try a lot of things and see what survives. Many of these side projects won’t make it; they won’t be able to find users or their maintainers will move on to other things. In the meantime, this does mean a lot of crap gets shipped though. I’ve seen a number of “open source” projects on GitHub that sound cool, but with a week of commits from six months ago and nothing since.
But from this diversity, some will survive. These may turn into companies, side hustles, or just toy projects that make the lives of their creators or users better. I’ve also seen some very cool new ideas built by small teams that grew into bigger ones, and large projects being dramatically refactored in ways that never would have been tried before.
AI doesn’t make software better by default, but it lets us make more of it, with more variety — exactly the raw material evolution needs. We won’t know which of today’s experiments matter until long after they’ve shipped, but we can at least try to enjoy the process and create some experiments of our own.