Victor Taelin of Higher Order Company has some of the hardest computer science problems the LLMs most likely have never seen before and evaluated Gemini 3 on them. Here is his tweet reproduced almost in full.

Short Version

First of all: you've all seen the benchmarks, so I don't think you need me to judge this one. Still, based on my tests, this is as real as it gets, and I want to talk about it. This model outperforms GPT-5 Pro, Gemini 2.5 Deep Think, and everything else, on my hardest problems, by far.

It is the new SOTA at:
→ debugging complex compiler bugs
→ refactoring files without logical mistakes
→ solving difficult λ-calculus problems
→ ASCII art (it is almost decent now!)
→ Competitive Gen 3 OU (won't elaborate 😭)

It is still an LLM, though. It has similar failure modes, and is worse than Sonnet / GPT-5 in some scenarios.

It seems very bad at:
→ inferring intent
→ not going overboard
→ one-shot vibe coding
→ creative writing
→ health questions

Also, I suspect this checkpoint isn't the best Google has.

Now, on to a complete, manually typed Gemini 3 overview.

Long Version

1. Vibe [...]

---

Outline:

(00:25) Short Version

(01:53) Long Version

---

First published:
November 18th, 2025

Source:
https://www.lesswrong.com/posts/N7oRkcz3PrNQSNyw9/victor-taelin-s-notes-on-gemini-3

---

Narrated by TYPE III AUDIO.

“Victor Taelin’s notes on Gemini 3” by Gunnar_Zarncke

LessWrong (30+ Karma)