got tired of onboarding my own AI agents to my own project every session. built a service that remembers everything for them. alpha is open
https://t.co/qEeaYHLWiV
#buildinpublic
- Claude, can you see the global CLAUDE.md rules? Read it.
- Yes, pulling MCP
- Just read the rules, I want to see what actually there
- Ok, MCP disabled in every future work.
...Fucking hell
#Claude
quick opus 4.7 test: threw it at my 10-agent pr-swarm on the same PR I reviewed yesterday with 4.6. went from 23 findings to 28. same PR, same skill, just the model swap
https://t.co/ewyLbO4Ewj
#ClaudeCode#buildinpublic
@chrislaupama been using codex and claude code for a while now. was on cursor about 8 months ago but the actual usage you get for the price isn't even close
so tired of "π¨ BREAKING" and "this changes EVERYTHING" and then it's some completely basic agent feature. is there genuinely no other way to get people's attention on here?
#AI
@pinaleontas that's what I'm building though. curated tweet discovery by topic, then AI that helps you say what you actually mean but better. not auto-generated slop, your real thought just polished
@ansubkhan for me it's all random honestly. sometimes claude tries to do some unhinged stuff, sometimes codex just fails at basic common sense. i use both π€·ββοΈ