Issue 10 · 11 September 2026
The Signal
Ten AI stories from the week, ranked by evidence rather than noise.
The three that led the week
Seven more worth your time
04OpenAI puts Pro subscriptions on hold due to Astra demandOpenAI has temporarily disabled sign-ups for its $200-per-month Pro plan. It's done this because demand for the new Astra model has strained its infrastructure.Heat 72.905OpenAI Launches the Agents API in Public Beta, Putting the Codex Harness Behind One API CallOpenAI has released the Agents API in public beta, putting the Codex harness behind one API call. The service gives developers access to the infrastructure that's running Codex.Heat 72.606Meta’s AI agent Muse is now the No. 2 app in the USMeta's new AI agent app, Muse, has taken the No. 2 spot on the US App Store chart. It's a ranking based on data from Sensor.Heat 71.807On the Navier–Stokes Millennium Prize ProblemOpenAI has published a solution to the Navier–Stokes existence and smoothness problem. It's one of the seven Millennium Prize Problems.Heat 71.708Mathematicians want proof OpenAI didn’t use their workMathematician Andreas Thom has publicly accused OpenAI of unethical data practices regarding its recent breakthroughs in formal mathematics. Mathematicians want proof that OpenAI didn't use their work.Heat 70.409OpenAI Wants to Know if an AI Industry Slowdown Would Even Be LegalOpenAI has asked members of Congress whether coordinating an industry-wide slowdown on frontier AI development would be legal. It's trying to determine if such an industry-wide pause is even legal.Heat 67.910Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeekAnthropic released a report on Thursday detailing distillation campaigns from Alibaba, Moonshot AI, and DeepSeek. It alleges persistent distillation attacks by China-based AI companies, noting that the activity's escalated as competition intensifies.Heat 67.4How the heat score works
Each story is scored out of 100: editorial significance contributes 55 points, recency 25, topic momentum across the week 15, and a disclosed frontier-model launch signal 5. Near-duplicate coverage is collapsed before the final ten are published.
Tested AI toolkits, made by us
One prompt, one job, done properly. We run every toolkit nine times before it goes on sale, have a rival AI grade the results, and ship the scorecard inside the download.
See the toolkitsNotion Template Architect
For the 9pm spreadsheet copying. Describe your workflow, get the Notion system.
Content Engine
For the blank LinkedIn box. A writing loop that edits itself before you see it.
Tested 9 times before sale · scored 9/10 by a rival AI. Our own products; some tool links on this site earn us a commission.
AI Maestro · The Signal, every Friday · all editions