Books

Science Catch-Up: When Science Finally Meets Reality

Science Catch-Up: When Science Finally Meets Reality

Sixteen chapters on the limits of the scientific method: what it costs to wait for institutional consensus, and how to judge knowledge that has not been blessed yet.

How I wrote it →

Papers

  • Preprint arXiv (cs.LG, cs.AI, eess.SY) · 2026

    An Omitted Mode Is a Rare Rule: The Sampling-Verification Danger Law in Continuous Code World Models

    Javier Aguilar Martín

    If a synthesised world model omits a rare failure mode, sampling will almost never catch it: N test rollouts miss an event of probability r exactly (1-r)^N of the time. Passing the tests certifies consistency with the sample and nothing more.

  • Preprint arXiv (cs.AI, cs.LG) · 2026

    When a Verified World Model Still Loses: Play-Adequacy vs Prediction-Accuracy in LLM-Synthesized Code World Models

    Javier Aguilar Martín

    A language model can write a game's rules as executable code, score high on predicting the next state, and still lose. Accuracy is not adequacy for planning.

  • Preprint arXiv (math.RA) · 2024

    The Derived Deligne Conjecture

    Javier Aguilar Martín, Constanze Roitzheim

    Brace algebras on operads give derived A∞-algebras a conceptual home, and with it new, rigorous versions of the Deligne conjecture.

  • PhD thesis University of Kent · 2023

    The Derived Deligne Conjecture

    Javier Aguilar Martín

    Doctoral thesis, supervised by Constanze Roitzheim. The work the preprint above came out of.

Teaching

  • Universidad de las Hespérides 2026–

    Lecturer

    Algebra & Geometry — 6 ECTS first-year core course, shared by the BSc in Maths & Data Science and Maths & Philosophy.

  • King's College London 2020–2025

    Graduate Teaching Assistant

    Undergraduate mathematics.

  • University of Kent 2020–2024

    Graduate Teaching Assistant

    Undergraduate mathematics, alongside the PhD.

  • KeepCoding Summer 2024

    Bootcamp instructor

    Statistical methods in R.