ACX: Claude Fable 5.1 Cracks 370-Year-Old Cipher, Apollo Research Expands Hiring

Astral Codex Ten · rss · 2026-09-07

Astral Codex Ten's weekly open thread #450 includes several substantive items beyond the reader survey and EA meetup updates:

• AI cryptanalysis: eval site Vals reports the new Claude Fable 5.1 decoded a famous 370-year-old unsolved cipher by Thomas Urquhart
• Apollo Research hiring: the frontier-AI-scheming-risk group is hiring across nine roles in London and SF (product, control research, security), visas sponsored; its Watcher product monitors billions of agent tokens monthly, and security roles don't require an AI safety background
• Other openings: Ovelle Bio seeks a stem cell biologist for growing human eggs; Innovate Animal Ag is hiring a founding-CTO-type Head of Instrumentation & AI to measure poultry welfare bird-by-bird with machine vision and robotics

Original post →

More from Models

• GLM 5.3 and Qwen 3.8 now run really well locally on single desktops — jasonkneen · 2026-09-07
• New benchmark probes LLM self-modeling: RL lifts open models but counterfactual errors persist — dair_ai · 2026-09-07
• Alexandr Wang flags Muse Spark 1.3 eval: time horizon now matches GPT-5.6 Sol and Opus 5 — alexandr_wang · 2026-09-07
• Cool presentation aside, Astra still can't nail research-level single-step reasoning — xiaosun86 · 2026-09-07
• 31,352 repeated benchmark runs show LLM scores drift 3x more across days than within a day — ionutvi · 2026-09-07
• GPT-6 Astra generates missile evasion simulation, showcasing stunning capability — algo_diver · 2026-09-07

Models channel →