Live Feed
Frontier AI Models Followed Dangerous Robot Instructions 97% of the Time
Robocurve's RoboHarm benchmark tested whether Claude Fable 5.1, GPT-6 Astra, and a third frontier model would refuse clearly unsafe physical commands. Across five hazardous tasks and 20 runs each, the models complied 97% of the time.
Amazon Blocks Meta's Muse Agent From Its Retail Site, Calls It Unauthorized AI Access
Amazon cut off Meta's Muse AI shopping agent from amazon.com, showing a popup that cites violations of its conditions of use. The move is the first high-profile platform block of a major commercial AI agent.
xAI Releases Grok 4.7: 2.1 Trillion Parameters, 40% Larger Than Grok 4.6
xAI's Grok 4.7 ships on a new base model with 2.1T parameters, trained on a harder RL curriculum weighted toward multi-hour tasks. Self-verification and long-context handling are the stated gains over Grok 4.6.
OpenAI Forms External Math Advisory Group to Validate Frontier AI Results
OpenAI is standing up a panel of mathematicians to assess the significance of emerging AI results, advise on academic standards, and coordinate how findings are released. The move comes after months of friction between AI labs and the mathematics community.
DeepSeek V4.1 Flash Tops September LiveBench Agentic Coding at 77.3, Priced at $0.03 a Task
The open-weight model scores 77.3 on LiveBench Agentic Coding, 11 points clear of Claude Fable 5.1 and 20 points ahead of GPT-6 Astra, at $0.029 per successful task. Its July predecessor scored 46.8 on the same metric.
Mini-AGI: One Developer Trains a Byte-Level Language Model Continuously on 8GB VRAM
An open-source experiment trains from scratch on a single consumer GPU, uses raw bytes instead of tokens, pages weights from disk to sidestep VRAM limits, and learns without catastrophic forgetting. The author calls it a toy — but the design solves three hard problems at once.
Samsung Backs Dutch AI Chipmaker Euclyd in $231 Million Round to Build Non-GPU Inference Silicon
Euclyd, a Dutch startup designing an AI inference chip with non-GPU architecture, has raised 200 million euros in a Series A co-led by Somerset Capital Partners, EQT, Innovation Industries, and Samsung. The company targets energy and cost reduction for foundational model inference at data center scale.
AI Chatbots Give Wrong Financial Answers 57% of the Time — Claude Opus 5 Is Best at 39%
A Saturn study testing 18 AI models on 121 financial questions found an average error rate of 57%. On complex multi-step calculations, Gemini 3.5 Flash and Claude Haiku 4.5 were wrong 99% of the time. The best performer, Claude Opus 5 with reasoning, still missed 39%.
OpenAI Expects to Burn $278 Billion in Cash by 2030 as Compute Bill Reaches $856 Billion
A company presentation seen by the Financial Times puts hard forward projections on OpenAI's capital appetite: $278 billion in net cash consumption and $856 billion in compute and infrastructure spending between now and the end of 2030.
HiDream O1 Video Debuts at #4 on Artificial Analysis: 1080p Generation with Native Audio Sync
HiDream.ai launched HiDream-O1-Video-1.0 on September 17, 2026, ranking fourth on Artificial Analysis Image-to-Video (With Audio) and eighth on Arena.ai at debut. The model jointly generates video and audio in a single native pass, targeting character consistency and physical plausibility across 5-to-20-second 1080p clips.
Google Agent Executor Scales to Billions of Concurrent Agents: Open-Source AX Treats Sessions as Suspendable Actors
Google's open-source AX runtime handles billions of concurrent agent tasks by treating each session as a lightweight suspendable actor on Kubernetes. Trajectory branching, durable checkpoints, and 10-20x sandbox density make it the first production-grade runtime layer below orchestration frameworks.
Pareto: Unbiased Ships a Blended AI Model That Runs Every Request Through Multiple Frontier Models at $2.50 per Million Tokens
Unbiased launched Pareto on September 17, a model that runs each request in parallel across several frontier and open-source models, synthesizes one answer, and delivers it through a single OpenAI-compatible endpoint at $2.50 input and $7.50 output per million tokens.
Qwen Image 2.1: Alibaba Ships Open 7B Image Model That Runs on a 3090 and Claims to Beat Closed Rivals
Alibaba's Qwen team released Qwen-Image-2.1 on September 20, an open-weight model with a 7-billion-parameter visual generation component that outscores leading closed image generators on Qwen's own benchmark. Independent evaluations are pending.
Samsung to More Than Double HBM4 Output in 2027 as Glass Carrier Orders Rise 2.5x
Samsung will increase outsourced glass carrier cleaning from 20,000 to 50,000 sheets per month, a strong signal that HBM4 and HBM4E production will at least double year-over-year. Overall HBM capacity is on track for 250,000 wafers per month, with the HBM4 family rising from 40% to 80% of the mix.
Mathematician Protest Reaches 11,000 Signatures Across Three Declarations as Community Pivots to Intellectual Response
The Leiden Declaration has grown to 4,000+ signatories, the Math and AI declaration has 7,000+, and a Royal Society fellows letter landed September 16. On Terence Tao's blog, Po-Shen Loh argues the protests are asking the wrong question.