HackerTrans
トップ新着トレンドコメント過去質問紹介求人

mmmehulll

no profile record

投稿

Discussion: Shared state breaks most distributed multi-agent systems

1 ポイント·投稿者 mmmehulll·6 か月前·0 コメント

Show HN: Ctrl – execution control plane for AI agents

github.com
2 ポイント·投稿者 mmmehulll·6 か月前·0 コメント

コメント

mmmehulll
·先月·議論
this is really cool. I didn't know we had these
mmmehulll
·先月·議論
love this. would be cool if we can see and perform all kinds of banking txns on this. Think ledger but all in one card. Super cool. Even cooler would be card to card money transfer without use of swipe machines
mmmehulll
·2 か月前·議論
why not cursor? cursor seems pretty decent at what it does and i have had very few issues since i started using it about a couple of months back. But yes, frontend burns through tokens, else no issues with backend and devops
mmmehulll
·2 か月前·議論
I agree. A month back, I was brainstorming ideas with AI a lot, and it took me a while to figure this out: I was essentially outsourcing my thinking. Honestly, I felt extremely dumb after the sessions and kept blaming myself for becoming dumb, but I stopped using AI, went back to my non-AI routine, and ideas started popping back up.
mmmehulll
·4 か月前·議論
What makes LLM bias reinforcement worse than a regular echo chamber is the confidence. A friend who agrees with you might say "yeah maybe", an LLM constructs a well-reasoned, authoritative-sounding case for whatever you've implied you believe.
mmmehulll
·4 か月前·議論
That’s really interesting. The “one or two examples + good error messages” part feels especially important. It suggests the limiting factor may be less finetuning and more whether the model is given a tight representation and a feedback loop it can recover from.
mmmehulll
·4 か月前·議論
This is really interesting, I feel if llms can respond at existing without finetuning, it can be huge.