@ericzakariasson Appreciate you sharing a list of actual loops! I've been scratching my head trying to think of loop usecases but I realize now that I was assuming loops had to be infinite. This list with bounded loops is much more practical.
Any suggestions for a markdown database for agent context? I want a central place for context that both agents can use and non-technical human users can view/modify.
@mntruell This is why I'd love to see a duplex voice mode in Cursor. Back and forth conversation with an orchestrator that manages agents under the hood.
@MartinTale@shipordie_ I like the aesthetic of the first few but the plain white color of the last one. Here's an example though it might be over-detailed
easier said than done of course... More often you'll end up in the weeds on how specific features should work, where to position UI elements, edge cases, etc. Trying to figure out how to overcome this.
Information density is a critical but non-obvious lever for building with AI. If you can concisely articulate the essence of your system, AI agents can bootstrap the full spec and self-orchestrate the development.
and this is coming from someone who's been using Claude models consistently since Claude 2... Sonnet 3.5 was my absolute favorite for a very long time and Opus 4.5 was another step level change. But Opus 4.7 and 4.8 have been a strange shift..
This echoes my experience with Opus 4.7 and now 4.8. It's so adversarial, insisting on challenging basic non-controversial statements and second-guessing everything. Very edgy and kind of paranoid.
Opus 4.8 is a very strange model. Clearly Anthropic tried to improve honesty, which is commendable. However, the model's curiosity (already worse in 4.7) degraded further. Result is a judgmental personality + sycophancy + sooo much hedging. Basically the opposite of Opus 3.