Claude cracks NIST curve seeds & Mercor maps AI benchmark gaps - Hacker News (Oct 17, 2026)

Support The Automated Daily directly: 15 free episodes a day, 5 editions, 3 languages, about $4 a day to produce. A coffee keeps it going: theautomateddaily.com/support Please support this podcast by checking out our sponsors: - KrispCall: Agentic Cloud Telephony - https://try.krispcall.com/tad - Effortless AI design for presentations, websites, and more with Gamma - https://try.gamma.app/tad - Prezi: Create AI presentations fast - https://try.prezi.com/automated_daily Today's topics: Claude cracks NIST curve seeds - Anthropic says Claude found that the seeds for the NIST elliptic curves P-256, P-384, and P-521 came from the C library generator lrand48, seeded with Unix timestamps. The finding weakens long-standing ECDSA backdoor theories. Mercor maps AI benchmark gaps - Mercor's Benchmark Map finds that AI benchmarks cover only a small share of O*NET occupations and tasks. About 90% of US working hours fall on tasks with little or no benchmark coverage. Claude Code rethinks code review - Anthropic's Claude Code team rebuilt code review around AI verification, automated bug and security review, risk-based human approval, and fast rollback. Merged pull requests more than doubled while reverts fell by over 40%. MLX update for Mac Studio - Apple's latest MLX release benchmarks large language models on M5 Ultra Mac Studio. It also supports running a single model across several Macs using Thunderbolt RDMA, for private local inference. PS5 turned into LLM server - PS5LM runs llama.cpp on a jailbroken PlayStation 5. It serves GGUF and multimodal models through OpenAI-style and Anthropic-style APIs on a home network, without the cloud. DeepSeek V4.1 Flash snapshot - DeepSeek's V4.1 Flash October snapshot appears on OpenRouter. It is a sparse mixture-of-experts model with a 1M-token context window and image input, aimed at coding and computer-use agents. Transformers as associative memory - An explainer treats MLP layers and self-attention as associative key-value memories. It shows how interference limits recall and how attention capacity grows with context length. JPEG XL versus AVIF showdown - Jake Archibald compares JPEG XL and AVIF for web images. AVIF wins for lossy photos and screenshots, while JPEG XL leads in lossless compression. Dune's Butlerian Jihad revisited - A look at the Butlerian Jihad, Dune's revolt against thinking machines. It explains why the franchise has Mentats instead of AI and why the story resonates in today's AI debates. - Claude Solves