PeopleofSuperintelligence

Researcher

Nate Rush

As of October 2026, Nate Rush is chief technology officer of METR, a research nonprofit that measures whether and when AI systems might threaten catastrophic harm to society. He co-authored HCAST in March 2025, a benchmark of 189 machine learning, cybersecurity, software engineering and reasoning tasks. A randomized trial he co-authored in July 2025 found that experienced open-source developers took 19% longer when they used early-2025 AI tools. Later he co-authored METR work on an expenditure horizon measure of AI agents' optimization ability, on whether discovery has accelerated, and on a per-action monitor for safer evals.

Now at
METR
Works
5
Citations
27
h-index
3

Publication numbers from OpenAlex, as of 15 October 2026.

Timeline

  1. March 2025

    The HCAST benchmark

    He co-authors HCAST in March 2025, a benchmark of 189 machine learning, cybersecurity, software engineering and reasoning tasks.

  2. July 2025

    A developer productivity trial

    The randomized trial he co-authors in July 2025 finds that experienced open-source developers take 19% longer when they use early-2025 AI tools.

  3. December 2025

    Measuring long software tasks

    He co-authors Measuring AI Ability to Complete Long Software Tasks, a paper in Advances in Neural Information Processing Systems 38 from the conference held in December 2025.

  4. March 2026

    Test-passing pull requests

    A METR note from March 2026, which he co-authors, finds that roughly half of test-passing SWE-bench Verified pull requests written by recent AI agents would not be merged into main by repository maintainers.

  5. July 2026

    The expenditure horizon measure

    He co-authors the July 2026 post that proposes expenditure horizon, a measure of AI agents' optimization ability, illustrated on the NanoGPT speedrun.

  6. August 2026

    Checking discovery rates for acceleration

    A note he co-authors with Tom Cunningham in August 2026 looks for slope changes in discovery metrics, to test whether discovery has accelerated.

  7. September 2026

    A per-action blocking monitor

    The September 2026 note he co-authors describes implementing and evaluating a basic per-action monitor for safer evals.

  8. As of October 2026

    Chief technology officer at METR

    As of October 2026, he is chief technology officer of METR, a research nonprofit that measures whether and when AI systems might threaten catastrophic harm to society.

Sources and credits

Photo: We have no published photo of Nate Rush yet. Send us a free-licensed photo.

  1. [2503.17354] HCAST: Human-Calibrated Autonomy Software Tasks arxiv.org. Checked 15 October 2026.
  2. Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity - METR metr.org. Checked 15 October 2026.
  3. Measuring AI Ability to Complete Long Software Tasks - proceedings.com proceedings.com. Checked 15 October 2026.
  4. Many SWE-bench-Passing PRs Would Not Be Merged into Main - METR metr.org. Checked 15 October 2026.
  5. Expenditure Horizon: Measuring Optimization Ability, with an Application to NanoGPT - METR metr.org. Checked 15 October 2026.
  6. Have We Seen an Acceleration in Discoveries? - METR metr.org. Checked 15 October 2026.
  7. Implementing and Evaluating a Basic Per-Action Monitor for Safer Evals - METR metr.org. Checked 15 October 2026.
  8. Nate Rush - METR metr.org. Checked 15 October 2026.