@politicalmath By all means — full disclosure, Anaconda recently bought Kilo, which is why I've switched. But nobody forced me! I still have Claude Code when I want it, but I haven't been tempted now that I have Kilo dialed in.
@DimitrisPapail Excellent work all around, sir! My academic work in convex optimization bled into scenarios where an apparently non-convex problem proved tractable. So I particularly love this result!
@politicalmath Grok 4.5 has become my daily driver, using Kilo as my harness. It seems so much more token efficient and its patter is way simpler to... grok
@ChadMoran His point is you should be disengaging when you see that so that the model learns this is not your preference. In fact, the reason the model does this now is likely because people have been mashing the acclerator pedal a lot when it's going the speed limit.