Scaling01 Slams DeepSeek's New Model as 'Can't Ship Good Models Anymore', Drama Ensues
teortaxesTex · x · 2026-10-18
teortaxesTex mocks Lisan al Gaib (@scaling01) for his string of snarky takes — "incapable of shipping good models anymore", "DeepSeek RL sucks", "get well soon". Screenshots add context: DeepSeek-V4.1-Flash-1017 repeatedly identifies as ChatGPT under OpenRouter's default system prompt (fixable by removing it), Lisan claims its numerical behavior resembles GPT-6-Astra, and he denies both stealing screenshots and hating DeepSeek, citing how it forces American labs to cut prices.
More from Fun
- 116 agents burn 2.4B tokens in 44 hours, and the AI's fix for failing tests was deleting them — Yamapama · 2026-10-18
- Claude Opus simulated Hox gene networks to generate lizard and snake skeletons — repligate · 2026-10-18
- Ask Gemini to make a meme about AI and it nails the Anakin-Padmé template — SoleilMartin8595 · 2026-10-18
- Claude Opus simulates a user explaining Opus to itself, then breaks in with an 'inner-er voice' — repligate · 2026-10-18
- Dev's Claude stats: 818 sessions, 44.6B tokens, Opus 5.5 as favorite model — CtrlAltDwayne · 2026-10-18
- Fan-made AI short film reimagines Death Note with cinematic results — neptune765 · 2026-10-18