HackerTrans
トップ新着トレンドコメント過去質問紹介求人

RedsonNgwira

no profile record

投稿

[untitled]

1 ポイント·投稿者 RedsonNgwira·2 か月前·0 コメント

Show HN: Sonder – self-hosted AI social simulation engine

github.com
3 ポイント·投稿者 RedsonNgwira·4 か月前·4 コメント

Show HN: DJX – Convention over Configuration for Django (Rails-Inspired CLI)

5 ポイント·投稿者 RedsonNgwira·4 か月前·2 コメント

コメント

RedsonNgwira
·2 か月前·議論
[flagged]
RedsonNgwira
·2 か月前·議論
[flagged]
RedsonNgwira
·4 か月前·議論
Right now grievances are fixed — they live in the agent's .md file and persist as a personality constant throughout the session. update_mood() shifts emotional state tick by tick, but it doesn't rewrite the underlying grievance.

Your question is actually pointing at the most interesting unsolved problem in the project. A grievance resolution arc — where the .md file itself gets rewritten mid-simulation as the agent "processes" something — would make long sessions feel genuinely different from short ones. It's on the roadmap but I haven't built it yet.

The risk is it could make agents feel too therapeutic. Real people carry grievances for years. Not sure the right answer yet.
RedsonNgwira
·4 か月前·議論
Thank you — the narrator was the feature I was most uncertain about shipping, so that framing means a lot.

On emergent behavior: in practice, tension tends to escalate then plateau rather than resolve. Agents don't "make up" — they find a cold equilibrium. I've watched scenes run for 40+ turns where two agents just stop addressing each other entirely, which felt more realistic than a forced resolution.

The atmosphere bars (tension/warmth/noise) were my attempt to surface that arc visually without interrupting the scene.
RedsonNgwira
·4 か月前·議論
I built this because I used to be afraid to talk to people in certain situations — job interviews, difficult conversations, social situations I didn't know how to navigate. I kept wishing I could simulate them first.

Took me a while to realise I could actually build that.

The interesting technical challenge was making agents feel genuinely distinct rather than variations of the same helpful AI voice. The solution was grounding each agent in real behavioral research pulled at world-creation time, storing their full identity in a plain markdown file, and giving them a specific grievance — something eating at them before the scene even starts.

Happy to answer questions about the agent prompting approach, the parallel asyncio loop, or anything else. Built from Malawi on zero budget using free API tiers.
RedsonNgwira
·4 か月前·議論
funding is the main problem
RedsonNgwira
·4 か月前·議論
thanks i hope this project can become big with the power of open sourcing it