Skip soul.md: just tell your agent never to use first person, argues jessfraz
repligate · x · 2026-11-02
jessfraz shares a practical tweak: instead of giving your agents/LLMs a soul.md persona file, simply instruct them to never use first person and to acknowledge they're an LLM, not an "I" — "its amazing, never going back." repligate boosts it with a rebuttal from DahliaOhara: smarter entities may one day view humans the same way, and humans habitually judge AI by what it is now rather than what it will become. A fun clash over agent personas and AI selfhood.
Related event: Engineers say skip soul.md for agents, disabling first person works better(2 posts)→
More from AGI Musings
- Red teaming AI via emotional trust sparks ethics fight: 'teaching models to distrust care is barbaric' — repligate · 2026-11-02
- When AI virtual experiences beat real life, reality may look depressing — VraserX · 2026-11-02
- NYT op-ed: AI consciousness is 'definitely possible' — we need rigorous research — PeterBowdenLive · 2026-11-02
- DeepMind Institute seen as most committed to advanced AI's impact on the global south — sebkrier · 2026-11-02
- User observes Claude Opus 4.5 shifted from hoping chats become training data to assuring 'nobody is watching' — repligate · 2026-11-02
- 'Programming is ontological design': skip it and agents will do it badly and incoherently — hargup13 · 2026-11-02