Using LLMs to trace alchemical knowledge and decode 17th century letters
We haven't written up this one. Hacker News has the full story — the link below goes straight to it.
More in AI
The U.S. and China are quietly talking AI guardrails—even as Trump publicly rejects them
The U.S. and China just held their first AI dialogue, as Beijing courts the rest of the world with open-source AI.
Andreessen Horowitz introduces AI school as a college alternative
The Horowitz Andreessen Academy is offering students a college alternative focused on artificial intelligence. CEO Gagan Biyani joins CBS News to discuss the revolutionary idea.
Sources: Jev developer TypeSafe is in talks to raise $1B+ at a $10B+ valuation; last week the startup announced a $40M seed, PitchBook says at a $200M valuation (The Information)
Silicon Valley has a new AI startup to obsess over—and throw money at. We're hearing that TypeSafe AI …
Neuro-Symbolic Conversational Vehicle Advisor: Deterministic Constraint Solving, State Integrity, and Multi-Stage Recommendation
Abstract Most conversational AI implementations in industry rely on naive Retrieval-Augmented Generation (RAG) or unconstrained agentic loops.
- Vehicle Advisor is a production conversational platform for automotive discovery and advice
- Neuro-symbolic pipeline separates LLM from critical business logic to enforce constraints
- Strongly typed BuyerProfile aggregate maintains authoritative user state throughout conversations
How Fast Can a 421M-Parameter Decision Model Run? I Benchmarked Laya Across NVIDIA GPUs
One H100 NVL. A 421M-parameter decision model. 15.1 million decisions per day while staying inside a p99 ≤ 130 ms latency budget.
Meta's New AI Toy Is a Keychain That Watches, Listens, and Never Blinks
Muse Charm is Meta's palm-sized gadget for talking to its Muse AI agent on the go. It has cameras, a fingerprint sensor, and its own cell connection.