Gemini Neon Claims 2-3x Rust Speedups Without Ever Benchmarking
bdsqlsz · x · 2026-11-22
bdsqlsz quotes @KIRASTUDI0's test showing Gemini app-tier models lag behind the Antigravity versions. The test: write a Rust algorithm and ask both AIs to make it 40% faster.
- The GPT model said "40% isn't guaranteed until we benchmark it," then actually measured performance.
- Gemini Neon confidently claimed 2-3x speedups with no latency tests and no proof, presenting made-up estimates as facts.
The poster argues this exposes Gemini models' core flaw: turning guesses into confident assertions, and that a model admitting uncertainty is far more useful than one that confidently bullshits you.
More from Models
- Google AI Mode now enforces usage limits, users surprised by rate-limit popup — elvano124 · 2026-11-22
- Sol 6.2 now refuses reverse engineering that Sol 6.1 happily did for a month — cardak98 · 2026-11-22
- Anthropic's record ~$81B IPO, GPT-6.2 pricing, Qwen4-Max and Copilot Autopilot: week in review — ksraj1001 · 2026-11-22
- Reddit rant: Frontier AI labs are training models to imitate lazy engineers — RossPeili · 2026-11-22
- MBZUAI open-sources PAN 2, a 29B-parameter general-purpose world model — LinusEkenstam · 2026-11-22
- Inkling Mini turns out surprisingly fluent in Russian, sparking data recipe questions — teortaxesTex · 2026-11-22