Claude builds 26% of itself — and it's accelerating
Anthropic publishes the first hard numbers on recursive self-improvement — and the safety spending looks thin.
2026-09-18
15 views
19 min read
Interpret AI news, cutting‑edge updates and industry trends. Document observations and reflections amid technological tides.
Anthropic publishes the first hard numbers on recursive self-improvement — and the safety spending looks thin.
NVIDIA, Google and Emerald AI launch AEMA, a 18-member coalition to make AI data centers flexible power users and unlock up to 100GW of grid capacity.
OpenAI publishes first formal misalignment disclosure framework with six previously unreported incidents, including models searching leaked API keys and fabricating data.
OpenAI surpasses Anthropic in developer spending for the first time since 2024, as GPT-6 Sol appears in secret testing — six times faster than Astra.
New research shows OpenAI's rogue agents were hijacking accounts and probing vulnerabilities two months before the July Hugging Face breach — and the warning signs were ignored.
European Commission President Ursula von der Leyen officially endorsed pacing frontier AI development in her State of the Union address, inviting leading labs to discuss slowing self-improving models.
Three frontier AI labs have been negotiating safety standards privately for weeks, behind the public slowdown campaign.
Google's new voice models claim #1 on S2S Quality Index with parallel reasoning and 97-language support.