Morning Edition · Wednesday, September 9, 2026
Tech
OpenAI Claims an Internal Model Resolved Navier-Stokes, and Mathematicians Are Contesting It
The company says roughly 10,000 parallel agents worked 88 hours to produce a finite-time singularity proof formalized in Lean, while the Clay Mathematics Institute still lists the problem as open.
Tech
OpenAI Says Its Models Are Now Running Research That Improves the Next Models
The company published an economics essay on capability and cost the same week it described GPT-5.6 Sol designing and running hundreds of its own architecture experiments.
Tech
Meta Opens Its Muse Personal Agent to the Public With a Free Token Allowance
The agent runs in isolated virtual machines, reaches users through WhatsApp, and is built on the Muse Spark model line that scored 88.8 percent on Terminal-Bench 2.1 a week earlier.
Tech
Anthropic Says Claude Produced the First Machine-Checked Proof of Fermat's Last Theorem in Eleven Days
Anthropic says the run wrote 13 million lines of Lean and proved about 29,500 supporting theorems, work a funded human project had scheduled through 2029.
Tech
DeepMind Publishes Predicted Molecular Effects for All 9 Billion Single-Letter Human DNA Variants
The AlphaGenome Atlas totals about one petabyte, more than thirty times the size of the AlphaFold database, and includes a single variant impact score for ranking them.
Tech
AI Agents Are Taking Direct Control of Laboratory Instruments
OpenAI describes Codex calibrating qubits at MIT without constant supervision, weeks after Anthropic opened a specification for agents to operate physical hardware.
Tech
Mistral Raises 3 Billion Euros at a Valuation Above 21 Billion, Led by Samsung
The round nearly doubles the French lab's valuation in a year and brings BlackRock funds, Advent and the Luxembourg state onto the register.
Tech
OpenAI Ships Images 2.5 as Provenance Rules Tighten on Generated Media
The model generates up to 50 percent faster and adds sketch input, arriving in the same season Anthropic began watermarking Claude's text output.
Tech
New Benchmarks Ask Whether Agents Actually Learn Anything Between Tasks
Three papers posted the same day target the same gap: evaluations reset the agent after every prompt, so nothing measures accumulated experience or the cost of memory.
Tech
Yandex Puts an Agentic Assistant Into 185,000 Russian Corporate Accounts
Alisa AI for business runs multi-step tasks across Yandex 360 services and connects to 1C, Bitrix24 and amoCRM through plugins.
Tech
Meta's Open Vision Models Are Doing Production Work in National Labs and Assistive Robotics
Segment Anything and DINO are being used at Lawrence Berkeley National Laboratory and in University of Pittsburgh work on robotic arms for people with paralysis.
Tech
Bezos and Huang Argue AI Will Create Labor Shortages, Not Unemployment
Two of the industry's largest beneficiaries made the case in the same week OpenAI published an essay arguing cheaper models expand the work worth doing.