Morning Edition · Tuesday, November 10, 2026
Tech
Anthropic ships Claude Fable 5.2 at a 20 percent lower price and keeps Mythos 5.2 behind vetted access
Fable 5.2 lists at $8 per million input tokens and $40 per million output tokens, down from $10 and $50, and Anthropic's own tables show Mythos 5.2 at 89.6 percent on a CVE-rediscovery test, against 78.3 percent for Mythos 5.1.
Markets
Anthropic heads toward a Nasdaq listing as OpenAI routes ChatGPT Enterprise through AWS procurement
Anthropic's S-1 shows first-half 2026 revenue of $16.29 billion, up from $1.16 billion a year earlier, while OpenAI now lets companies buy ChatGPT Enterprise through their existing Amazon Web Services billing.
Tech
OpenAI's Control Room can halt any registered research job within 35 seconds and fails closed
Each training, evaluation or data job gets an identifier, and the inference engine checks with Control Room before serving each request, so if the service goes down, inference stops.
Tech
Terence Tao calls for a "Math 2.0" after OpenAI's release of more than 700 machine-generated manuscripts
Tao says some of the papers contain strong ideas but are badly written, miss attributions and sometimes contain errors, and the dispute has moved from whether models can prove theorems to who verifies and credits the output.
Tech
Claude helped resolve the first 3D structure of a prokaryotic vault particle, Anthropic says
Anthropic's lab reports that a purification step that usually takes about three days took about 90 minutes, and it also discloses a protocol error by Claude that cost roughly two weeks.
Tech
Meta's SAM 3 and DINOv3 run inside Energy Department beamline analysis and a robotic wheelchair project
Meta says Berkeley Lab cut a grapevine micro-computed tomography segmentation task from about a month of expert work to about 15 minutes, and a University of Pittsburgh team is adapting the same models to run on the wheelchair itself.
Tech
NVIDIA uses CoRL to move simulation-trained robotics from locomotion to fine assembly
NVIDIA and partner labs presented simulation work on aligning gears, connecting cables and assembling small parts, tasks where contact physics, not gait control, limits sim-to-real transfer.
Tech
NVIDIA plans investments in Korean AI startups at a government forum that included domestic chip rivals
South Korea's science ministry hosted more than 20 startups at COEX, and Korean reports say they included the neural processing unit (NPU) designers DeepX, Rebellions and FuriosaAI, which compete with NVIDIA in inference.
Tech
Public agent skill packs from Matt Pocock, Cloudflare and Obsidian's CEO spread across Claude Code and Codex
A Telegram tally puts Pocock's skills repository at 319,000 GitHub stars, and every install loads third-party instructions that run with the agent's own permissions.
Tech
Moonshot publishes Kimi K3.1 Pro weights as tiny Qwen-based models win narrow tasks
Kimi K3.1 Pro is reported as a 2.8-trillion-parameter mixture-of-experts model with 104 billion active parameters and a 1M-token context, and an 800-million-parameter fine-tune of Qwen3.5 is claimed to beat far larger models at generating Bash commands.
Tech
NormGym tests whether agents can infer unwritten social rules through interaction, not just state them
The new arXiv benchmark targets the ability to pick up socially enforced expectations in unfamiliar communities and adjust behavior, which most existing norm tests do not measure.
Tech
Google AI Studio now embeds live Google Maps in apps that Gemini generates from a text prompt
A prompt such as a 3D tour of New York landmarks now returns a working map app, which extends the Maps grounding tool Google made generally available in the Gemini API in October 2025.