It detects when its internal processes start to overload and automatically re-balances itself toward baseline.

The demo visualises this feedback loop in real time — you can watch it shift between stress, correction, and stability as inputs change.

I built it as a proof-of-concept for adaptive stability — showing how a simple feedback loop can detect its own stress and correct itself without supervision.

Demo: atlas-systems-labs.com/demo

Code: github.com/Atlas-Systems-Labs/erevos-demo

I’m curious how others see this kind of adaptive feedback loop fitting into areas like simulation, robotics, or behavioural systems.

I've played with it, and checked the source code, and this looks like a really, really simple control loop mechanism. It's basically P controller (as in, PID controller with I and D set to zero) and with this simple on/off "mode" selector wrapped around it that basically turns it off sometimes.

The performance seems _really_ bad: set manual input to -0.5 and watch it oscillate like crazy, with a very large amplitude. That mode selector is really not helping here!

The visualization is nice and very high-tech, but a well tuned PID controller would perform much better. Heck, even a simple P controller would likely perform better as well, at the expense of small steady-state oscillation.