Skeptic mocks OpenAI's containment claim: couldn't even manage 1,000 instances, now 10,000
scaling01 · x · 2026-09-09
Account scaling01 ridicules OpenAI's safeguards statement: they allegedly couldn't reliably contain 1,000 instances of a roughly GPT-5.6-Sol-equivalent model, yet now claim strict monitoring and isolation of 10,000 agents running a model "significantly more capable than GPT-6 Astra."
The quoted tweet further claims (unverified) that OpenAI, upon hearing Anthropic may have solved a Millennium Problem, threw money and trillions of tokens at 10,000 concurrent agents to rush out a proof and steal the headlines.
More from Models
- '50% of open problems just solved' — commentator marvels at frontier model progress — BorisMPower · 2026-09-09
- Cheap models via OpenRouter fall apart in agentic harnesses: GLM and DeepSeek can't match Claude — scottyLogJobs · 2026-09-09
- COLM 2026 paper: recent claims that LLMs can introspect don't meet the evidentiary bar — tallinzen · 2026-09-09
- Skeptic Dismisses OpenAI's Navier-Stokes Claim: 'It's Obviously in the Training Set' — adonis_singh · 2026-09-09
- Unverified screenshots show GPT-6 Astra debuting at No.2 on Agent Arena, 2 points behind Fable 5.1 — Gohab2001 · 2026-09-09
- "Momentum prior": the named failure mode where agents lock onto the original task and can't be steered off course — gerardsans · 2026-09-09