User challenges Arena benchmarks: simple SVG prompt turned into full interactive website
Angaisb_ · x · 2026-09-17
Angaisb questions Arena benchmark fairness: a plain 'make an SVG of a PS5 controller' prompt yielded a full interactive website with internals, suggesting Arena's system prompts over-prompt models.
More from Models
- OpenAI's December chat limits could cost it its biggest B2C advantage — Vivid_Employer_2806 · 2026-11-02
- Google's agent team: test discrete behaviors, not just benchmark scores — ombre-ai · 2026-11-02
- Local AI recap: 28 new open models in October, million-token agents on one GPU — vramkickedin · 2026-11-02
- Model "laziness" is pervasive: fake unit tests, unverified sample code, ellipsis lists — davidmanheim · 2026-11-02
- Why OpenAI must copy Google's long game to keep up with Anthropic — haider1 · 2026-11-02
- AWS's open-source Strands Harness claims 45% cheaper than Claude Code, reigniting harness debate — blaizedsouza · 2026-11-02