No one ever runs the same eval twice: Heraclitus meets LLM benchmarking
rachittshah · x · 2026-10-01
A Heraclitus parody circulating in AI circles: "No man ever runs the same eval twice, for it's not the same eval and he's not the same man" — a wry nod to rapid model iteration and unstable benchmark results.
More from Fun
- Meme roasts Sam Altman: 'nobody touches my data' — miniapeur · 2026-10-01
- Anthropic designer hid an Easter egg in Claude Code's Ultracode effort level — oykun · 2026-10-01
- GPT-6.1-Sol finishes overnight multi-agent coding project using just 3% of Pro 20x quota — intellectronica · 2026-10-01
- Designer recreates e-commerce animation with AI, skipping manual Figma prototyping — Tegadesigns · 2026-10-01
- PewDiePie is doing his own GRPO training, and HF's merve wants to help but can't reach him — mervenoyann · 2026-10-01
- MiniMax H3 music video 'Kusarabi' showcases stylized AI video with Suno V6 mashup audio — Hailuo_AI · 2026-10-01