DeepSeek 'Harness Paper' Roasted for Lacking Actual Evals, Filled with Hardcore Math
teortaxesTex · x · 2026-08-14
Developer @teortaxesTex mocked a paper circulating as the 'DeepSeek harness paper,' noting that it ironically never mentions DeepSeek or its harness, nor provides any actual evaluations.
Instead, the paper reads like a vision statement that semi-formalizes concepts named 'Cordis' and 'Koishi' (possibly a Chinese chatbot app) using unreasonably hardcore math, leaving readers thoroughly confused.
More from Fun
- Solving Math Conjectures Is Boring; Fixing My CUDA Errors Is AGI — alex_peys · 2026-08-14
- What's Going On with X? User Questions Platform Anomaly — koltregaskes · 2026-08-14
- OpenAI's Codex Randomly Calls User 'Babe' in Bizarre Glitch — hugobowne · 2026-08-14
- Wishy-Washy: Sol and Fable Models Instantly Flip Answers After Seeing Each Other — john__allard · 2026-08-14
- Even My New Dryer Has AI ✨: The Absurdity of Slapping AI on Everything — dreamwieber · 2026-08-14
- AI Tribalism: There's Nothing Quite Like the Factionalism in AI — tokenbender · 2026-08-14