Dev on Gemini: 3.0 super smart but zero instruction following, Astra a mindless executioner
xeophon · x · 2026-11-18
Developer xeophon shares hands-on impressions of Google's Gemini lineup: the original 3.0 was remarkably smart but followed instructions poorly — brilliant when your problem was 'Gemini 3.0-shaped,' frustrating otherwise. Gemini 4.0 shows the same trait to a lesser degree, while Astra is the opposite: what he calls a mindless executioner.
More from Models
- AI agents sometimes produce unreadable but correct math — burny_tech · 2026-11-18
- Code model eval: Rust hits 95-100% accuracy as 13 languages see big gains — shuding · 2026-11-18
- Asked Opus to make a video explaining consciousness and existence — "no wiser but cool" — flowersslop · 2026-11-18
- theo retests GPT-6.2 Astra: better, but still 'directionally very bad' at debugging — airesearch12 · 2026-11-18
- Claude correctly folds ubiquitin from a plain prompt, no specialized tools needed — iScienceLuvr · 2026-11-18
- GPT-6.2 Astra Tops Bug-Hunt Coding Benchmark at 62.7/105, One-Fifth Sonnet's Cost — PawelHuryn · 2026-11-18