I spent a few hours yesterday developing an evolutionary computation simulator.
It’s called EvoLab:
- EvoLab: Browser-based neuroevolution sandbox
Here’s the code project:
It’s your classic “evolve a creature for an environment” thing.
Here’s a screenshot:
Back in the day I tried to develop something like this for iphone and it took weeks just to get the physics and creature working.
The first working version of EvoLab was developed in a few minutes using Claude Opus. Then there were a few iterations of refining the user interface and evolutionary mechanisms.
This is crazy.
It’s exactly what I envisioned when I sat down and it “just works”.
I sat there for a long time playing with it and thinking. It felt hollow.
I guess what I wanted was to design and build it more than the final product.
This is fascinating. Perhaps most of my projects are like this. Motivated by a desire to build the thing rather than to have/use the thing.
The solution, I think, is to instantiate more whims and dig down into this further.
Build.