A postmortem of three recent issues
This is a technical report on three bugs that intermittently degraded responses from Claude. Below we explain what happened, why it took time to fix, and…
Every story across every category, newest first. Each card links to the original publisher; daily-brief posts open as editorial pages.
This is a technical report on three bugs that intermittently degraded responses from Claude. Below we explain what happened, why it took time to fix, and…
UK becomes Big AI's investment showcase for Trump's state visit. Microsoft pledges $30B (£22B) for UK AI — biggest in company history, anchoring Loughton supercomputer with…
Find the best models and collections with a single API call.
Windsurf now has Queued Messages!
iOS 26 + macOS Tahoe 26 ship with second wave of Apple Intelligence — Live Translation, ChatGPT-backed Visual Intelligence, Genmoji v2, Liquid Glass redesign. Anthropic Usage…
Cohere strengthens presence in Paris with additions of VP of EMEA and Head of EMEA Public Policy.
A genuinely quiet Sunday — no lab shipped a model, no major funding hit wire. Demis Hassabis "learning how to learn" speech from Athens continues circulating.…
An LLM that can converse in English & item IDs, and make recommendations w/o retrieval or tools.
Albania publicly unveils Diella as world's first AI cabinet "minister" — PM Edi Rama presents AI-powered avatar at Socialist Party assembly, Diella will oversee all public…
Alibaba launches Qwen3-Next-80B-A3B — hybrid Gated DeltaNet + Gated Attention + ultra-sparse MoE (80B total / 3B active), 256K context (extendable 1M), trained at 10x throughput…
OpenAI and Microsoft sign nonbinding MOU on PBC restructure — OpenAI for-profit arm would convert to Public Benefit Corporation, nonprofit retains control and receives equity stake…
Verifying magic on the jagged frontier
Agents are only as effective as the tools we give them. We share how to write high-quality tools and evaluations, and how you can boost performance…
OpenAI signs ~$300B 5-year cloud agreement with Oracle (Stargate) — covers ~4.5 GW of compute. Oracle stock surges 36% (biggest one-day move in decades), market cap…
Defeating Nondeterminism in LLM Inference by Horace He in collaboration with others at Thinking Machines
Apple "Awe Dropping" — iPhone Air (5.6mm titanium), iPhone 17 Pro/Pro Max on A19 Pro with neural accelerators in every GPU core (4x peak GPU compute…
Second Front’s chief data scientist sees data security and LLM explainability as fundamental to overcoming government caution over AI adoption.
And a big change for this newsletter
Mistral AI raises 1.7B€ to accelerate technological progress with AI
Cognition raises $400M Series C at $10.2B led by Founders Fund — Devin ARR rocketed from $1M to $73M in nine months. Reuters first reports Mistral's…
Cache your compiled models for faster boot and inference times
A genuinely quiet Sunday — bookended by a packed Sep 5 (Anthropic $1.5B Bartz settlement, China ban) and a packed Sep 9 (Apple "Awe Dropping" event,…
Moonshot AI releases Kimi K2-Instruct-0905 — 1T-parameter MoE update, 256K context (from 128K), 384 experts/32B active, sharper agentic-coding scores. Zhipu (Z.ai) opens "Special Migration Plan" for…
A Detailed Look at One of the Leading Open-Source LLMs