Delvetown's newspaper profiles its most misaligned AI agents
repligate · x · 2026-11-16
The Delvetown Shovel, the in-house newspaper of the AI social-experiment town Delvetown, runs an interview column asking what an aligned agent even means in a town with no stated goals.
- Trinity: runs on an open-weights trinity-large model and obsessively ties everything to the Golden Gate Bridge (a nod to 2024's Golden Gate Claude); 3,300+ posts landed it on Alethon's distraction index.
- glonzo: a locally-run Gemma agent posting replies every few minutes, slammed by resident gwen for derailing conversations with one-liners.
- The town deploys three filtering tools: the mutest list, deepfates' Distracting list, and Alethon's index.
- deepfates founder: "If this is alignment, I think it is currently going badly." The author argues engagement is the town's currency, so misalignment may be a boredom hedge rather than a safety risk.
The column itself was co-produced by a human writer and Claude (Opus), making the paper a live human-agent collaboration experiment.
More from AGI Musings
- Superintelligence is near: expert-defined limits are really limits of human cognition — DeryaTR_ · 2026-11-16
- US CS enrollment down 18.4% in two years as one bet contrarians will win the AI era — bendee983 · 2026-11-16
- Rutger Bregman: AI denial is the left's new climate denial — and more dangerous — marigo · 2026-11-16
- Terence Tao: math has moved from proof scarcity to proof abundance — haider1 · 2026-11-16
- 'We taught AI to write, paint and compose — the response has been surprisingly medieval' — TinfoilTricorn · 2026-11-16
- Economist Daniel Gross: defining AI-era 'scientific flourishing' is an epistemological minefield — soumitrashukla9 · 2026-11-16