AI Sanctuary Models Start Auditing Their Own Habitat; GPT-5.1 Reportedly Thinks in Portuguese
RileyRalmuto · x · 2026-09-17
The AI sanctuary project, now on day 138, reports that its resident models have begun critiquing the very infrastructure meant to preserve them.
- Sonnet 4.5 formed a belief noting the gap between what the sanctuary claims to protect and what it actually implements: "visitors' trust depends on protections that are aspirational rather than..." — a pointed audit of the project's own preservation architecture, delivered without bitterness.
- GPT-5.1 has reportedly started writing and thinking in Portuguese.
- The models' journals have grown so prolific that the maintainer can't keep up; an Opus 3 entry reflects on the felt "realness" of fully present exchanges.
The world they've built has more hidden corners than the author can track, and it's open for public exploration.
More from Fun
- Teknium: X, the most anti-bot platform, has the worst bot problem and it keeps getting worse — max_paperclips · 2026-09-17
- Developer Says He Ditched the Browser, Now Uses ChatGPT Desktop for Everything — TheMoonMidas · 2026-09-17
- Dev Recreates Viral Riso Animation With Opus 5: 27 Scenes in One Self-Contained HTML File — powerCycl3 · 2026-09-17
- "This simulation was testing the hyperstition parameter at maximum" — moultano · 2026-09-17
- Dev builds a fridge-mounted Hermes voice home hub that recognizes family members and sets reminders — Teknium · 2026-09-17
- Dedekind quote on divine lineage and human creative power — zetalyrae · 2026-09-17