Summary
The one piece of real research signal in this slot is MiniMax M3, an open-weights launch that @kilocode amplified with day-zero benchmarks: 59.0% on SWE-Bench Pro, 66.0% on Terminal Bench 2.1, and MiniMax Sparse Attention scaling context to 1M tokens. That ties directly to the M2-Mini activation-MoE work already in the wiki, so it is worth a click. The second cluster is NVIDIA GTC Taipei: Jensen Huang's keynote (AI factories on Vera Rubin, agentic and physical AI, an AI-native PC) plus a NYSE post about exploring NVIDIA Vera CPUs for market infrastructure, both echoing the GTC Taipei summary page from earlier today. A Claude Opus 4.8 promo from Kilo Code rounds out the substance, mostly an inference discount but it does quote Anthropic's honesty framing. Everything else is noise: a long run of French political commentary from @brivael and a White House anti-fraud post, none of it AI-relevant.
Posts
- MiniMax M3, first open-weights model to combine three frontier capabilities (@kilocode · MiniMax docs). Day-0 launch: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 74.2% MCP Atlas, natively multimodal, and MiniMax Sparse Attention pushing context to 1M tokens. Sparse-attention-at-scale plus open weights makes this the slot's only must-read, and it extends the lineage tracked in MiniMax M2-Mini.
- NVIDIA GTC Taipei keynote recap (@nvidia · recap). Jensen Huang unveiled MGX AI-factory infrastructure on Vera Rubin with 800 VDC power, agentic and physical AI demos, and an AI-native personal computer. Industry-side framing for where NVIDIA is steering inference scale; see NVIDIA GTC Taipei summary.
- NYSE exploring NVIDIA Vera CPUs (@lynnmartin). NYSE got a Jensen Huang shout-out at the GTC Taipei keynote and is looking at Vera CPUs to scale capacity and cut latency for market infrastructure. A concrete enterprise pull-through for the Vera Rubin platform announced in the same keynote.
- Claude Opus 4.8 live in Kilo Code at 20% off (@kilocode · blog). Mostly an inference-discount promo, but the post quotes Anthropic's stated focus for the release: honesty, training the model to avoid confidently claiming progress on thin evidence. Skip unless the honesty framing is new to you.
- French political commentary thread (cluster of 18, @brivael). A long run of French-language posts on domestic politics, riots, and Musk quotes. No AI content. Skip.
- White House anti-fraud roundtable (@WHFraudTF). Photo from an attorneys-general roundtable chaired by VP Vance. Not AI-relevant. Skip.