# anthropics/original_performance_takehome

Anthropic's original performance take-home, now open for you to try!

Repository: https://github.com/anthropics/original_performance_takehome
Canonical: https://ross.abutalabs.com/products/original_performance_takehome
Language: Python
License Family: other
Last push: 2026-01-22T01:11:08+00:00

## Health v2 (maintenance only)
Score: 44/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 63, release rhythm 35, longevity 16
- inputs: {"age_days": 226, "days_push": 224, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4122, forks 943 (observed 2026-08-28T04:08:35.994020+00:00)

## What it is
Anthropic's open-sourced performance engineering take-home challenge, where you optimize a simulated machine's code from a slow baseline (232,000 cycles) toward record cycle counts. It doubles as a benchmark comparing human and LLM optimization ability and as a recruiting channel for Anthropic.

## Use cases
- practice low-level performance optimization on a real challenge
- compare my optimization skills against Claude Opus 4.5's benchmark results
- try Anthropic's hiring take-home with unlimited time
- benchmark how well an AI agent performs at performance engineering
- learn cycle-level optimization and debugging with provided tools
- submit a sub-1487-cycle solution to impress Anthropic recruiters

## When to choose
- you want a self-contained, well-instrumented performance optimization challenge
- you're interested in testing LLM coding agents on hard optimization tasks
- you want a shot at getting noticed by Anthropic's recruiting team

## When to avoid
- you need production software or a reusable library
- you want a maintained tool with a license or ongoing support
- you're looking for a general-purpose benchmarking framework

## Facets
- artifact type: learning-resource
- maturity: stable
- function: benchmarking, developer-tools, simulation
- domain: performance, developer-tools, tutorials
- platform: python, cli, cross-platform
- tags: coding-challenge, performance-optimization, take-home-exercise, recruiting, llm-benchmark, low-level-optimization

## Member repositories
- anthropics/original_performance_takehome (main) score 44

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:35.994020+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:23:05.777312+00:00, confidence not recorded.
  - readme: https://github.com/anthropics/original_performance_takehome (fetched 2026-08-28T04:08:35.994020+00:00, sha 0fdda30d7acd)
- Data as of 2026-08-30T08:39:29.467469+00:00.