Fetching from the wire…
01
02
03
04
05
06
07
08
Policy2026-09-10 · source-backed
Announced September 9. Christiano originated RLHF, most recently advised CAISI, and has publicly put substantial probability on catastrophic outcomes from AI. Appointing him during an active Senate investigation is either a real governance change or the most legible possible signal of one, and there's no way to tell which from outside yet.
Each link below shares sources, entities, or timing with this story.
David Vélez and Robin Vince joined both the OpenAI Foundation and OpenAI Group PBC (OpenAI). Both skew heavily toward finance and regulated-industry governance rather than AI research or product. Read alongside this month's IPO reporting, the board composition looks like publi...
Researchers reveal that Direct Preference Optimization implicitly operates over a full preference graph, meaning it extracts more signal from existing datasets than anyone realized. Practical implication: your existing RLHF data may be more valuable than you think.
Pacing the Frontier went public July 28 with signatures from OpenAI, Anthropic, Google DeepMind, Meta, Microsoft, Mistral, and Thinking Machines, asking the U.S. government to lead an international effort on the technical and governance tools needed to deliberately pace automa...
Moonshot AI released Kimi K3, a sparse mixture-of-experts activating 16 of 896 experts per token. That's about 1.8% of the pool live at any moment, with a 1M-token context window and native vision. Two new architectural pieces show up: Kimi Delta Attention and Attention Residu...
The 2026 consensus: exhaust prompting and retrieval before touching weights, and when you fine-tune, use a thin LoRA/QLoRA adapter paired with retrieval, not full fine-tuning. DPO is the default over RLHF when you have preference pairs. Decision rule: RAG for knowledge that ch...
ARC-AGI-2 Reasoning Race. Gemini 3.1 Pro hit 77.1% and Claude Opus 4.6 hit 68.8% — both more than doubling their predecessors' scores in a single generation. The entire field moved ~10 points over the prior two years; each model jumped 30-40 points in one release. The top thre...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.