Gemini 3.8 Flash Divides Opinion: 'Joke' on Benchmarks, But Solving Hard PHP WASM Kernel Problems
MickeySteamboat · x · 2026-09-09
A spat over Gemini 3.8 Flash: oleks01 cites Terminal-Bench results calling the model a joke and warning not to be fooled by flashy charts; MickeySteamboat fires back blaming user error and a bad kernel setup, claiming the model is solving some of computer engineering's hardest PHP WASM kernel problems. The gap between benchmark scores and real-world experience is on display again.
Related event: Gemini 3.8 Flash Divides Opinion: Strong Benchmarks, Poor Real-World Use(2 posts)→
More from Fun
- Employee vs Freelancer vs Solo AI Founder: Three Very Different Income Curves — saibharadwaj · 2026-09-09
- Meme pokes fun at Anthropic doom warnings every time OpenAI ships a better model — StewartalsopIII · 2026-09-09
- Slop existed before AI: a reminder as fears of AI slop mount — TejasKumar_ · 2026-09-09
- 'AI will take your job' — the AI in question: a glorious onboarding fail — emeka_boris · 2026-09-09
- Remastering a 40-Year-Old Anime with GPT-Image 2.5 and Local MiniMax H3 — heliumcraft · 2026-09-09
- Michael Levin's Controversial Platonic Space Paper Clears Peer Review — drmichaellevin · 2026-09-09