Ashby gave the example of an autopilot - if you flipped the yoke controls so that up was down and down was up, a traditional autopilot would go into a positive feedback loop and crash, but the Homeostat would adapt to the new dynamics.
He postulated that this was a model for some systems in the brain and perhaps all learning (he wrote an entire book about it) and there's some evidence he was right. If you put goggles on someone that flips their vision upsidedown, they adjust after a few days... and have to readjust again when you take them off! (This was a real experiment.) YouTuber SmarterEveryDay found he could learn to ride a "backwards bike," but it destroyed his ability to ride a normal bike. You may have experienced this first hand if you've ever played a video game where the controls were flipped temporarily: it's disorienting, sure, but you quickly adjust.
Because of these phenomena, I think Ashby was right about the Homeostat being a useful model of the brain. It explains why so many apparently contradictory diets "work" - simply making a major changes resets the homeostat in your brain where it may settle into a better calibrated equilibrium eventually.
It implies a simple strategy: are you happy? If yes, we're done. If no, change something, and return to the question. (I've seen this as a flowchart online somewhere but can't find it now.)