Google researchers find a way to keep self-improving AI agents from memorizing their tests
Newly verified: The RRSI authors report gains of up to 4.7 points across five out-of-distribution benchmarks.
SourceVane · Practical AI
Source-reviewed AI changes published from 2026-09-28 through the following Sunday.
2026-09-28–2026-10-04
The week’s source-linked changes, without manufacturing a trend where the evidence does not support one.
Newly verified: The RRSI authors report gains of up to 4.7 points across five out-of-distribution benchmarks.
Newly verified: The Decoder reports that GPT-6 Astra leads the accompanying benchmark with up to 53 percent reconstruction accuracy.
Newly verified: The Decoder reports that Anthropic's revenue increased twelvefold in 2025 to nearly $4.6 billion.
TechCrunch AI reports that before OpenAI launched its new AI agent, Dots, on Tuesday, Elon Musk's xAI had already acquired the domain name "dot.com," which now redirects to the Grok chatbot download page.
TechCrunch AI reports that OpenAI says GPT-6.1 Sol delivers significant improvements over GPT-6 Sol across complex professional tasks, including code writing and debugging, document understanding, and executing multi-step business workflows.
What happened next · API prices · Decision guides · Latest AI coverage · Follow updates