

4·
3 days agoSeems to be a common impression. Based on my experience: what about improving the tooling itself so you don’t need to guide as much? Then having it produce an artifact that the process worked? Like a report. Then what about comparing 3 way with other algos/approaches? Meanwhile what about setting up that new project or fixing a pet peeve? How about 3 tasks at the same time with 4D chess? Sounds snarky but that’s what I’ve been doing, plus a lot of free time to deal with personal things or readings. One more example: if it takes 10m can it take 5? LLM to the experiments. Is there a new research paper to apply to that data? LLM to find out

You can help it, good repo-level agents directives, personal directives, tooling checks, CI by another model, self-review. Bunch of stuff that comes packaged in those toolkits. But I think the power is in the personal customization not in those canned solutions. So you ask your best assistant what you asked me and how to ensure rigour for publication. Not that at I have experience with that. That’s tough. Sometimes its like “if you what you did is right then run this old data and prove it still works” kind of opportunities and we don’t notice.