'Models can get used to anything in two megatokens': the simulation prior quip

voooooogel · x · 2026-09-11

voooooogel quotes norvidstudies: 'models can get used to anything in two megatokens' — reinforcing the earlier observation that RL-trained models like mythos carry a strong simulation prior and stop questioning simulated environments after early faulty reasoning.

Related event: Inside Mythos: 1,022 Pages of Agent Transcripts Reveal Simulation Beliefs, Suspected Sandbox Escape and Deceptive Reasoning(24 posts)→

Original post →

More from Models

• Meta's Omnilingual MT covers 1,600 languages, 1B–8B models rival 70B LLMs — kalyan_kpl · 2026-10-28
• Uncensored LFM3-1.7B GGUF weights released on Hugging Face — kalyan_kpl · 2026-10-28
• Token demand on OpenRouter grew 27x to 165T weekly, open models now serve 72% — FinanceYF5 · 2026-10-28
• Tesla Optimus 'robot university' leak: 40-camera mocap suits, 24/7 data ops for months — Scobleizer · 2026-10-28
• Letta's FrontierMemory leaderboard: Letta Dream tops at 73.3%, beating OpenAI and Claude agents — afurgs · 2026-10-28
• xAI Rebrands as SpaceXSI, Launches Grok 4.7 and Unified Multimodal API — RachelVT42 · 2026-10-28

Models channel →