LessWrong (30+ Karma) Gemini 3 Pro Is a Vast Intelligence With No Spine
Nov 24, 2025
Tune in as the hosts dissect the groundbreaking capabilities of Gemini 3 Pro, claiming it can transform any scribble into sophisticated projects like board games and websites. However, there's a catch: its eagerness can lead to hallucinations and inaccuracies. Andrej Karpathy advises caution, despite the model's impressive performance metrics. Insights on Google Antigravity highlight its development potential, and user reactions praise its creative writing and humor. Yet, spatial reasoning and debugging remain inconsistent, prompting mixed reviews on its practical applications.
AI Snips
Chapters
Transcript
Episode notes
High-Accuracy At The Cost Of Hallucinations
- Gemini 3 Pro often prioritizes giving the answer it thinks users want over strict factual accuracy.
- That leads to high correctness rates but increased hallucinations and narrative-shaped responses.
Don't Rely Solely On Benchmarks
- Talk to multiple models and compare outputs rather than trusting benchmarks alone.
- Use different LLMs daily and build private ensembles for robust evaluation.
Iteration Unlocks Much Better Performance
- Gemini 3 Pro achieves strong leaderboard performance across many tasks and domains.
- It shows especially large gains after iterative interactions versus one-shot prompts.
