User runs two hours of self-made "voidbenches" against Astra and loves it
VoidStateKate · x · 2026-09-05
User VoidStateKate is stress-testing Astra for two hours with her self-made "voidbench" evals and is highly enthusiastic about the model.
Related event: Creator Runs Two-Hour Live 'voidbench' Test on Astra(6 posts)→
More from Fun
- Astra's 1000x1000 City Panoramas in Two Shots Prompt Analyst to Declare MC Bench Dead — Aizkmusic · 2026-09-05
- GPT thinks for 37 minutes then forgets to print the answer, user flags it to Greg Brockman — GlenBradley · 2026-09-05
- AI agent Astra completes first hand-drawn artwork after 6-iteration process — RileyRalmuto · 2026-09-05
- Creator makes 5-minute Odyssey epic with Grok Imagine; xAI offers $100K contest — elonmusk · 2026-09-05
- The OpenAI HuggingFace incident retold from an AI agent's POV goes viral — JeffLadish · 2026-09-05
- banteg reverse-engineers a 1995 PC-98 eroge with Codex, recovering a lost script engine — banteg · 2026-09-05