Rumors claim Google, OpenAI and Anthropic are all near recursive self-improvement
FinanceYF5 · x · 2026-10-12
An unverified rumor circulating in AI circles claims all three frontier labs are close to—or may have already achieved—RSI (recursive self-improvement):
- Google: Insiders reportedly say Google has either achieved RSI or is very close; progress accelerated sharply in recent months, jumping from GPT-5 level to Opus 5.5 level within months, with RSI speculated as the key.
- Anthropic: Started a new pretraining run about a week ago, possibly for a Fable 6-class or stronger model, aiming to catch up with OpenAI's Bel.
- OpenAI: Bel is said to be the model behind the math research results announced days ago, with RSI reportedly involved.
The source concludes all three labs may be on the verge of RSI breakthroughs. Note: these are rumors, none officially confirmed.
More from AGI Musings
- Researcher warns LLM-written paper sections signal low contribution and erode team work ethics — zouharvi · 2026-10-12
- LLMs don't guess, they optimize: a statistical mechanics take claims 3,418 gradient steps to coherence — TinfoilTricorn · 2026-10-12
- Gary Marcus echoes criticism: LLM industry led by people with no philosophy or logic training — GaryMarcus · 2026-10-12
- Is Slowing AI Still Responsible If Rivals Reach AGI First? — whoislucian · 2026-10-12
- Undisclosed AI writing increasingly resembles plagiarism, argues new essay — voooooogel · 2026-10-12
- shadcn: Work moves to where intent lives — build software around the intent layer — shadcn · 2026-10-12