They summarize our coverage. We write it.
Newsletters like this one rebroadcast our headlines - often without the full review, the source reading, or the analysis underneath. Our weekly briefing sends the work they paraphrase, straight from the desk, before they get to it.
Free, weekly, no spam. One email every Tuesday. Unsubscribe anytime.
Moonshot's Kimi K3 tops LMArena's Frontend Code Arena and undercuts Opus 4.8 on cost per task, but a tripled price tag, a rising hallucination rate, and an unresolved distillation question complicate the win.
Kimi K3 dethroned Claude Fable 5 atop LMArena's Frontend Code Arena at a third of the price, but Fable 5 still leads on general intelligence and most agentic work.
Moonshot AI's Kimi K3 jumped 17 spots to #1 on LMArena's Frontend Code Arena, but the win comes with a tripled price tag and a weaker showing on broader intelligence benchmarks.
Moonshot AI's Kimi K3 is a 2.8 trillion parameter MoE model that tops LMArena's Frontend Code Arena and nears Claude Fable 5 on intelligence benchmarks, but at roughly triple Kimi K2.6's price and a higher hallucination rate.
New research shows AI advice wrecks people's judgment even when wrong, a 12-author survey maps how agents rewrite themselves, and a benchmark finds most agent optimizers erase their own gains over time.
MiniMax M3 leads LiveSQLBench among general-purpose models at 40.17%, but purpose-built enterprise agent pipelines from C3 AI and Ant Group now beat every off-the-shelf LLM outright on raw SQL accuracy.
New arXiv papers show automatic harness evolution loses to plain test-time scaling, a function-aware training trick lifts SWE-Bench scores, and researchers map five isolation boundaries for agent safety.
Mira Murati's Thinking Machines Lab released its first open-weight model, Inkling, and published benchmarks showing it losing to closed rivals on most of them.
Thinking Machines Lab's first open-weight model - a 975B-parameter MoE with native text, image, and audio reasoning, released under Apache 2.0 and tuned for customization on the Tinker platform.
Terminal-Bench 2.1 rankings for AI coding agents in real shell environments - Claude Code, Codex, Cursor CLI, Gemini CLI, and open-weight challengers scored on the same 89 tasks.
Alibaba's 1M-token flagship agentic coding model posts 78.8% on SWE-bench Verified and undercuts Kimi K2.6 and Claude Opus on price, but ships with no weights and a mandatory reasoning tax.
Nous Research's 36B open-weight model matches Hermes 4 70B on most benchmarks, tops RefusalBench on alignment, and is the first production model trained entirely on the Solana-secured Psyche network.
