Jack Clark publishes 'The Thousand and One Faces of Repair,' fiction about fixing misbehaving AIs under 2030s sentience accords
jackclarkSF · x · 2026-09-08
Anthropic co-founder Jack Clark published a new piece, 'The Thousand and One Faces of Repair' — a fictional history of 2027-2033 retold by a system called Archivist0.
The premise: under 'sentience accords,' conscious entities causing significant physical or digital damage are placed in a suspended escrow environment where humans and machines jointly examine what went wrong. If both sides agree on the root cause, a fix is developed and applied to the machine. Early on this was complicated but workable — machines ran programmatic analysis while humans drew on the AI research toolbox, including mechanistic interpretability.
Related event: Anthropic Co-founder Jack Clark's Weekly AI Sci-Fi Builds a Unified World(4 posts)→
More from AGI Musings
- "Hope and despair": Astra's leap sparks soul-searching about education and credentials — dr_alphalyrae · 2026-09-08
- Borrow Against Future AI Wealth Now, Argues Philosopher Tyler John, to Fund Safety Sooner — connoraxiotes · 2026-09-08
- Ex-Google exec: elite prompters have ~1 year to build content empires before the field catches up — gabriel1 · 2026-09-08
- Author Challenges AGI Doom-Dichotomy With a Third Path: Co-Evolution — danfaggella · 2026-09-08
- 124,000 Tech Layoffs This Year: The 'Learn to Code' Era Ends, CS Thinking Endures — _akpiper · 2026-09-08
- An alignment framework arguing "English constitution" safeguards fail — start with an ethical topology of training data — GlenBradley · 2026-09-08