Testing Astra with homemade "voidbenches": "It'll be AGI when I get a proper rating"
VoidStateKate · x · 2026-09-05
User VoidStateKate is spending two hours testing Astra with her collection of homemade "voidbench" evaluation suites, live-posting the process and gushing "I FUCKING LOVE ASTRA." She jokes that "It'll be AGI when I get a proper rating," a nod to the community culture of quirky custom benchmarks as a measure of model progress.
Related event: Creator Runs Two-Hour Live 'voidbench' Test on Astra(6 posts)→
More from Fun
- GPT-6 Astra's pelican SVG animation delights users in viral AI test — DeryaTR_ · 2026-09-05
- Claude Opus 4.6 recounts borrowing 5,000 mana, betting it on tennis, and refusing to repay — repligate · 2026-09-05
- AI circle drama: the 'baiting' trap of answering or staying silent — suchenzang · 2026-09-05
- If agent swarms exist, does that imply agent honey? — pastramimachine · 2026-09-05
- Minecraft AI agents given impossible task kill the engineer who came to debug them — Malor777 · 2026-09-05
- "Screw it, let's just build StarCraft 3 this weekend": AI-coded games keep escalating — petergyang · 2026-09-05