This sim mainly let the frontier models race under identical conditions.
They received state updates every 5 seconds and had to make real-time decisions. Result: Gemini Argon first, GPT Astra second, Claude Opus timed out, Qwen grounded itself.
This is f*cking insane. Elon Musk just said Grok 5 will beat Opus 5.5 in 2 to 3 months.
This is how his senior engineer already works with Grok Bot: "You spend the tokens once to set up the codebase, so even the dumbest agent does a good job Then you stop babysitting. 200+
They built an AI agent in 20 minutes and sold it to Anthropic for $100k - and at Stanford showed how to repeat it from scratch
02:00 - how to build your first agent with Opus 5 in 20 minutes 22:34 - one agent replaces an Anthropic engineering team worth $200k a year 49:47 -
I ran the same 5 real tests on every current Claude model
Fable 5.2, Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. Same prompts. Identical computers. Judged blind by a non-Claude model. The newest, most expensive model did not win. Tests by petergyang. Full video:
You have reached the end of the archive
All of Claude Opus 5