Rumors claim Google close to cracking recursive self-improvement, Anthropic and OpenAI racing too
nikola_mr64990 · x · 2026-10-17
Unverified rumors from @LearnWithSubhan: Google insiders claim it has cracked or is near recursive self-improvement (RSI), jumping from GPT-5 to Opus 5.5 level in months; Anthropic started a pretraining run for Fable 6 or bigger; OpenAI's Bel model behind its recent math results also allegedly used RSI. All hearsay, unconfirmed by any lab.
More from Models
- Swiss team's open Apertus 1.5 adds multimodal input and 262K context, with all training checkpoints public — lmoroney · 2026-10-17
- DeepSeek Harness v0.2.1 nears release with major plugin system overhaul — teortaxesTex · 2026-10-17
- Argon impresses at non-coding tasks, claimed to outscore all OpenAI and Claude models — bindureddy · 2026-10-17
- Huawei details training DeepSeek V4 on Ascend 950 supernodes: 273K tok/s on 2 racks — teortaxesTex · 2026-10-17
- NVIDIA releases Deciffusion: 8B diffusion model for 24ms agent decisions — PavloMolchanov · 2026-10-17
- DeepSeek users spot possible V4.1-Flash-1017 refresh across its API — chemist_slime · 2026-10-17