From the archive
Monday, 5 October
Folded from 6 sources that day.
Sorted by quality and by how many sources covered it, the same for everyone.
-
Our approach to EU text provenance rules
OpenAI explains how it is approaching text watermarking under EU rules, covering where watermarks apply, how detection works, and why access to detection starts with researchers.
-
Import AI 475: Swarm scaling; Google DeepMind watermarks biology; and the AI science economy
Toby Ord models multi-agent swarms as a new form of inference scaling. Swarms use about twice the tokens for the same performance but finish in half the wall-clock time, with diminishing returns similar to human teams.
-
Data Engineering Weekly #290
Roundup with Raschka on LLM-based text classification without task-specific training, a decision-model approach to PII checks and LLM routing, and Uber Eats halving search latency by cutting redundant work.
-
Coding-agent spending by OpenAI researchers is doubling roughly every month
OpenAI data shows median daily coding-agent usage by researchers, at API prices, rose from under $1 in January 2026 to $601 by mid-August. Power users grew at a similar rate.
-
Building advertising for the way people use AI
OpenAI is adding a new visual ad format to ChatGPT, along with expanded measurement tools, attribution partnerships and brand suitability controls for advertisers.