PeopleofSuperintelligence

Researcher

Jean-Stanislas Denain

Also known as JS Denain

Jean-Stanislas Denain is a senior researcher at Epoch AI, where he leads the Insights team. After entering École polytechnique in 2017, he completed a Statistics PhD at UC Berkeley under Jacob Steinhardt in 2025, with a dissertation titled "Measuring Artificial Intelligence: Bottom-up and Top-down Approaches." His early papers with Steinhardt tested representation similarity measures, model visualization methods and how language models process false demonstrations. In November 2024, he co-authored FrontierMath, a benchmark of original, expert-vetted mathematics problems on which state-of-the-art models solved under 2% at release. At Epoch, his April 2026 report with Alexander Barry found strong evidence that AI capabilities had accelerated on three of four metrics, driven by reasoning models.

Now at
Epoch AI

Timeline

  1. 2017–2025

    Studies at École polytechnique

    He enters École polytechnique in 2017 through the MP computer science track of its entrance exam, joining the X2017 class. The school awards him its diploma in March 2025.

  2. 2020–2025

    Statistics PhD at Berkeley

    He enters the Statistics PhD program at UC Berkeley in 2020, advised by Jacob Steinhardt. In 2025, he graduates with the dissertation "Measuring Artificial Intelligence: Bottom-up and Top-down Approaches."

  3. 2021

    Testing representation similarity

    In 2021, he coauthors "Grounding Representation Similarity with Statistical Testing" with Frances Ding and Jacob Steinhardt. The NeurIPS 2021 paper checks whether measures such as CCA and CKA respond to changes in a network's functional behavior.

  4. 2022–2023

    Auditing visualizations and overthinking

    With Steinhardt in 2022, he tests whether model visualization methods can flag anomalously trained models, and finds that they struggle to identify subtle anomalies. A 2023 paper with Danny Halawi and Steinhardt describes "overthinking" and "false induction heads" in language models given incorrect few-shot demonstrations.

  5. November 2024

    Co-authoring FrontierMath

    In November 2024, he is one of the co-authors of FrontierMath, a benchmark of hundreds of original mathematics problems crafted and vetted by expert mathematicians. State-of-the-art AI models solve under 2% of the problems at release.

  6. October 2025

    Counting OpenAI's digital workers

    At Epoch AI, he co-writes "How many digital workers could OpenAI deploy?" with Anson Ho and Jaime Sevilla in October 2025. The essay estimates that OpenAI has the inference compute for tens of millions of digital workers on a narrow set of tasks.

  7. April 2026

    Measuring AI acceleration

    In April 2026, he and Alexander Barry publish "Have AI capabilities accelerated?", which finds strong evidence of acceleration on three of four capability metrics, driven by reasoning models. Also in April 2026, he and Yann Rivière estimate how fast production of humanoids, robotic arms and drones could scale.

  8. As of October 2026

    Leading Epoch's Insights team

    He is a senior researcher at Epoch AI, where he leads the Insights team. In September 2026, he joins Nathan Lambert on the Interconnects podcast to debate the US-China gap, jaggedness and the trajectory of AI.

Career

  1. NowEpoch AISenior Researcher

Education

  1. 2025École PolytechniqueDiploma
  2. 2025University of California, BerkeleyStatistics PhD

From the biography.

  1. Direct Form 990 pay in the latest returns of 39 nonprofits with AI people on this site is highest at $2.1M at the Allen Institute for AI (Ai2) (2024)Includes Epoch AI. IRS Form 990 compensation across 39 nonprofits with AI people on this site from 2012 to 2025: in the latest returns, Mozilla reports $6.1M including related-company pay (2024) and the Allen Institute for AI (Ai2) leads direct nonprofit pay at $2.1M (2024).
  2. 19% of people shaping AI who founded a dated organization have founded two or more, waiting a median of 7 years before a secondIncludes Epoch AI. Across 1,136 people shaping AI who founded or co-founded a dated company, standalone lab or nonprofit, 221 (19%) founded at least two organizations, accounting for 36% of all 1,421 founding events.
  3. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Epoch AI. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes University of California, Berkeley. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Jean-Stanislas Denain yet. Send us a free-licensed photo.

  1. Jean-Stanislas Denain - JORFSearch jorfsearch.steinertriples.ch. Checked 20 October 2026.
  2. Jean-Stanislas Denain | Department of Statistics statistics.berkeley.edu. Checked 20 October 2026.
  3. [2108.01661] Grounding Representation Similarity with Statistical Testing arxiv.org. Checked 20 October 2026.
  4. [2206.13498] Auditing Visualizations: Transparency Methods Struggle to Detect Anomalous Behavior arxiv.org. Checked 20 October 2026.
  5. [2307.09476] Overthinking the Truth: Understanding how Language Models Process False Demonstrations arxiv.org. Checked 20 October 2026.
  6. [2411.04872] FrontierMath: A Benchmark for Evaluating Advanced Mathematical Reasoning in AI arxiv.org. Checked 20 October 2026.
  7. How many digital workers could OpenAI deploy? | Epoch AI epoch.ai. Checked 20 October 2026.
  8. Have AI capabilities accelerated? | Epoch AI epoch.ai. Checked 20 October 2026.
  9. Jean-Stanislas Denain | Epoch AI epoch.ai. Checked 20 October 2026.
  10. Debating RSI, the US-China Gap, and Jaggedness with JS Denain of Epoch AI interconnects.ai. Checked 20 October 2026.