Skip to main content
Andrei Moise

On this page

  1. The search loop — branch → measure → prune → deepen
  2. Two lessons
  3. References
All notes
2 min readAgentic Workflows

Beating Anthropic's Original Performance Take-Home Challenge

1,361 cycles, down from a 147,734-cycle baseline, with no background in kernel optimization, SIMD, or compiler work.

Earlier this year, I took Anthropic's Original Performance Take-Home with no background in kernel optimization, SIMD, or compiler work.

Result: 1,361 cycles, down from a 147,734-cycle baseline (108.5×), passing all 9 published thresholds.

How? I asked questions, used agentic search, switched models, and applied different techniques.

The search loop — branch → measure → prune → deepen

Tree-of-thoughts-shaped exploration over candidate optimizations. Another idea began with self-consistency but evolved into exploration. Seven parallel sessions with the same prompt, then comparing investigations, branching on candidates, measuring cycles, and pruning failures.

I switched models when plateaued (Claude ↔ Codex ↔ external GPT analysis). Different training data and post-training will result in different blind spots.

I also used agentic search to find what others did, how and why.

Caught them cheating. Once outright test edits, once semantic cheating that only a clean-room re-test exposed.

Two lessons

Don't avoid problems where you lack expertise; AI can compensate. Be careful what you prune; don't outsource judgment, problem-solving, and understanding all at once.

We can outsource some of it, some of the time, not all of it at once. The skills we cut are our own.

References

Reach out

ai@andreimoise.com