The lead
SMART combines 200 data streams to flag congestion and suggest schedule changes. FAA staff review its advice and can reject it.
Prefer audio? The briefing has chapters and a full transcript.
The briefing
4 more stories
OpenAI called for US-led international technical standards on September 21. Its proposal covers common evaluations, human oversight and incident reporting.
OpenAI has proposed a policy, not an international rule. The company also says fully autonomous recursive self-improvement is not happening today. Its warning about losing control describes a possible future.
Read article →The Gates Foundation announced a 60-organization coalition on September 21. Over five years, the coalition aims to help an estimated 3.4 billion people use AI in their own language and voice.
The foundation is an interested participant. This is a target population, not a count of people already served. The coalition says its structure and governance will be developed over the next year.
Read article →Treasury Secretary Scott Bessent said on September 21 that senior US and Chinese officials would meet in Shenzhen in about two months. Reuters reported that he made the comments to CNBC.
This remains Bessent's account of the talks. The report does not show that either side has tested an incident line or settled the notification rules. Bessent said they still need to discuss the protocols.
Read article →AI's energy demand drew attention at New York Climate Week on September 21. AP reported disagreement over whether AI's future benefits would outweigh the burden from data centers.
Forecasts of cleaner computing do not measure today's emissions. Even a useful AI application may consume more resources than it saves.
Read article →Analysis
A reviewer needs evidence, time and the power to refuse. That is true for flight planners checking SMART forecasts and governments assessing an AI lab's claims.
The FAA can compare recommendations with actual delays and the choices staff made.
A lab can propose a test. Others still need a way to examine its results.
Language users need a say in data collection and a way to report answers that fail them.
The watchlist
Flight planning and AI standards put human oversight to a practical test.
Editorial direction, not a financial index. Each signal is tied to this edition’s reporting.
Edition validated · 5 stories · 8 unique sources