<?xml version="1.0" encoding="utf-8"?>
<?xml-stylesheet type="text/xsl" href="../assets/xml/rss.xsl" media="all"?><rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Vestigium (Publicaciones sobre AlphaProof)</title><link>https://jaalonso.github.io/vestigium/</link><description></description><atom:link href="https://jaalonso.github.io/vestigium/tags/alphaproof.xml" rel="self" type="application/rss+xml"></atom:link><language>es</language><copyright>Contents © 2026 &lt;a href="mailto:"&gt;José A. Alonso&lt;/a&gt; 
&lt;a rel="license" href="https://creativecommons.org/licenses/by-nc-sa/4.0/"&gt;
&lt;img alt="Creative Commons License BY-NC-SA"
style="border-width:0; margin-bottom:12px;"
src="https://i.creativecommons.org/l/by-nc-sa/4.0/88x31.png"&gt;&lt;/a&gt;</copyright><lastBuildDate>Thu, 01 Oct 2026 13:00:02 GMT</lastBuildDate><generator>Nikola (getnikola.com)</generator><docs>http://blogs.law.harvard.edu/tech/rss</docs><item><title>Readings shared April 4, 2026</title><link>https://jaalonso.github.io/vestigium/posts/2026/04/04-readings_shared_04-04-26/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;The readings shared in &lt;a href="https://bsky.app/profile/jalonso.bsky.social"&gt;Bluesky&lt;/a&gt; on 4 April 2026 are:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://leodemoura.github.io/blog/2026-4-2-why-lean/"&gt;Why Lean?&lt;/a&gt;. ~ Leonardo de Moura. #LeanProver #ITP&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.24823v1"&gt;A formalization of the Gelfond-Schneider theorem&lt;/a&gt;. ~ Michail Karatarakis, Freek Wiedijk. #LeanProver #ITP #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.23095v1"&gt;Formalizing Pick's theorem, efficiently&lt;/a&gt;. ~ Michael Eisermann. #LeanProver #ITP #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.26342v1"&gt;Lean on Vampire proofs (short paper)&lt;/a&gt;. ~ Jonas Bodingbauer, Márton Hajdu, Laura Kovács, Axel Polaczek, Michael Rawson. #LeanProver #ITP #Vampire #ATP&lt;/li&gt;
&lt;li&gt;&lt;a href="https://es.gizmodo.com/la-ia-acaba-de-hacerle-una-pregunta-incomoda-a-la-fisica-teorica-que-pasa-si-algunos-de-sus-resultados-mas-citados-nunca-fueron-tan-solidos-como-parecian-2000229508"&gt;La IA acaba de hacerle una pregunta incómoda a la física teórica. Qué pasa si algunos de sus resultados más citados nunca fueron tan sólidos como parecían&lt;/a&gt;. ~ Martín Nicolás Parolari. #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2604.01483v1"&gt;Type-checked compliance: Deterministic guardrails for agentic financial systems using Lean 4 theorem proving&lt;/a&gt;. ~ Devakh Rashie, Veda Rashi. #LeanProver #ITP #LLMs&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.28406v1"&gt;Physics as code: From scans to theorems with ITP APIs in SU(5) model building&lt;/a&gt;. ~ Sven Krippendorf, Joseph Tooby-Smith. #LeanProver #ITP #Physics&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.27007v1"&gt;Pairwise independence of representation, classification, and composition in finite extensional magmas&lt;/a&gt;. ~ Stefano Palmieri. #ITP #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.28636v1"&gt;Optimal bounds for an Erdős problem on matching integers to distinct multiples&lt;/a&gt;. ~ Wouter van Doorn, Yanyang Li, Quanyu Tang. #AI4Math #LeanProver #ITP #LLMs&lt;/li&gt;
&lt;li&gt;&lt;a href="https://hxrts.com/telltale/papers/paper1.pdf"&gt;Coherence for asynchronous buffered MPST: A mechanized metatheory&lt;/a&gt;. ~ Sam Hart Berman. #LeanProver #ITP #Physics&lt;/li&gt;
&lt;li&gt;&lt;a href="https://hxrts.com/telltale/papers/paper2.pdf"&gt;Computable dynamics for asynchronous MPST: Lyapunov descent, critical capacity, and decision procedures&lt;/a&gt;. ~ Sam Hart Berman. #LeanProver #ITP #Physics&lt;/li&gt;
&lt;li&gt;&lt;a href="https://hxrts.com/telltale/papers/paper3.pdf"&gt;Harmony from coherence in asynchronous MPST: A minimal erasure invariant for classical dynamics&lt;/a&gt;. ~ Sam Hart Berman. #LeanProver #ITP #Physics&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.25682v1"&gt;On the formalization of network topology matrices in HOL&lt;/a&gt;. ~ Kubra Aksoy, Adnan Rashid, Osman Hasan, Sofiene Tahar. #ISabelleHOL #ITP #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.20405v1"&gt;Putnam 2025 problems in Rocq using Opus 4&lt;/a&gt;.6 and Rocq-MCP. ~ Guillaume Baudart, Marc Lelarge, Tristan Stérin, Jules Viennot. #RocqProver #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://youtu.be/5fAWYqM9v8k"&gt;Representing graphs in Prolog&lt;/a&gt;. ~ Markus Triska. #Prolog #LogicProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://youtu.be/EjNalq6EihU"&gt;Encodings into the λ-calculus&lt;/a&gt;. ~ Kristopher Micinski. #LambdaCalculus #Racket #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://stopa.io/post/269"&gt;What Gödel discovered&lt;/a&gt;. ~ Stepan Parunashvili. #Logic #Math #Lisp #Programming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.technologyreview.com/2026/03/25/1134642/this-startup-wants-to-change-how-mathematicians-do-math/"&gt;This startup wants to change how mathematicians do math&lt;/a&gt;. ~ Will Douglas Heaven. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.29961v1"&gt;Short proofs in combinatorics and number theory&lt;/a&gt;. ~ Boris Alexeev, Moe Putterman, Mehtaab Sawhney, Mark Sellke, Gregory Valiant. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://mathscholar.org/2026/04/how-effective-are-current-ai-models-on-mathematical-research-problems/"&gt;How effective are current AI models on mathematical research problems?&lt;/a&gt;. ~ David H Bailey. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2603.26524v1"&gt;Mathematical methods and human thought in the age of AI&lt;/a&gt;. ~ Tanya Klowden, Terence Tao. #AI #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://abuseofnotation.github.io/category-theory-illustrated/06_type/"&gt;Category theory illustrated: Types&lt;/a&gt;. ~ Jencel Panic. #CategoryTheory #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/06/11-resena-de-leantutor-a-formally-verified-ai-tutor-for-mathematical-proofs/"&gt;Reseña de 'LeanTutor: A formally-verified AI tutor for mathematical proofs'&lt;/a&gt;. #ITP #LeanProver #Math #AIforMath #Teaching&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/06/13-resena-de-matp-bench-can-mllm-be-a-good-automated-theorem-prover-for-multimodal-problems/"&gt;Reseña de «MATP-BENCH: Can MLLM be a good automated theorem prover for multimodal problems?»&lt;/a&gt;. #AI #MLLMs #Math #ITP #IsabelleHOL #LeanProver #CoqProver #AIforMath&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/06/13-resena-de-mathesis-towards-formal-theorem-proving-from-natural-languages/"&gt;Reseña de «Mathesis: Towards formal theorem proving from natural languages»&lt;/a&gt;. #AI #LLMs #Math #ITP #LeanProver #AIforMath&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/06/15-hardest-problems-in-mathematics-physics-the-future-of-ai/"&gt;Reseña de «Hardest problems in mathematics, physics &amp;amp; the future of AI»&lt;/a&gt;. #ITP #LeanProver #AI #Math #AIforMath&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/06/15-hablemos-de-lisp/"&gt;Reseña de «Hablemos de Lisp»&lt;/a&gt;. #Lisp #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/07/01-diophantine-equations-over-z-universal-bounds-and-parallel-formalization/"&gt;Reseña de «Diophantine equations over Z: Universal bounds and parallel formalization»&lt;/a&gt;. #ITP #IsabelleHOL #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/07/27-solving-formal-math-problems-by-decomposition-and-iterative-reflection/"&gt;Reseña de «Solving formal math problems by decomposition and iterative reflection»&lt;/a&gt;. #AI4Math #LLMs #ProofAssistants #LeanProver #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/"&gt;Reseña de «The Infinity Project (How to use AI and mathematics to prove and improve science and security)»&lt;/a&gt;. #AI #Math #ITP #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/11/14-olympiad-level-formal-mathematical-reasoning-with-reinforcement-learning/"&gt;Reseña de «Olympiad-level formal mathematical reasoning with reinforcement learning»&lt;/a&gt;. #AI #Math #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/11/21-deepminds-latest-an-ai-for-handling-mathematical-proofs/"&gt;Reseña de «DeepMind’s latest: An AI for handling mathematical proofs»&lt;/a&gt;. #AI #Math #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/12/05-formalization-of-erdos-problems/"&gt;Reseña de «Formalization of Erdős problems»&lt;/a&gt;. #ITP #LeanProver #Math #Autoformalization&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/02/13-resena-de-machine-assistance-and-the-future-of-research-mathematics/"&gt;Reseña de «Machine assistance and the future of research mathematics»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/02/17-resena-de-mathematics-and-machine-creativity-a-survey-on-bridging-mathematics-with-ai/"&gt;Reseña de «Machine assistance and the future of research mathematics»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/02/27-resena-de-ai-that-can-prove-its-right-verification-as-the-missing-layer-in-ai/"&gt;Reseña de «AI that can prove it’s right: Verification as the missing layer in AI»&lt;/a&gt;. #AI4Math #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/03-resena-de-completing-the-formal-proof-of-higher-dimensional-sphere-packing/"&gt;Reseña de «Completing the formal proof of higher-dimensional sphere packing»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/09-resena-de-shaping-the-future-of-mathematics-in-the-age-of-ai/"&gt;Reseña de «Shaping the future of mathematics in the age of AI»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/10-resena-de-ai-for-mathematical-and-scientific-discovery/"&gt;Reseña de «AI for mathematical and scientific discovery»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/11-resena-de-mathematicians-in-the-age-of-ai/"&gt;Reseña de «Mathematicians in the age of AI»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/23-resena-de-coding-after-coders-the-end-of-computer-programming-as-we-know-it/"&gt;Reseña de «Coding after coders: The end of computer programming as we know it»&lt;/a&gt;. #AI #Programming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/24-resena-de-eval-awareness-in-claude-opus-46s-browsecomp-performance/"&gt;Reseña de «Eval awareness in Claude Opus 4&lt;/a&gt;.6’s BrowseComp performance». #AI&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/25-resena-de-formal-verification-and-ai-are-reshaping-mathematical-research/"&gt;Reseña de «Formal verification and AI are reshaping mathematical research»&lt;/a&gt;. #AI4Math #ITP #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/27-resena-de-in-math-rigor-is-vital-but-are-digitized-proofs-taking-it-too-far/"&gt;Reseña de «In math, rigor is vital&lt;/a&gt;. But are digitized proofs taking it too far?». #AI4Math #LeanProver #ITP&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/27-resena-de-can-machines-do-mathematics/"&gt;Reseña de «Can machines do mathematics?»&lt;/a&gt;. #AI4Math #LeanProver #ITP&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/27-impulsar-las-matematicas-guiando-la-intuicion-humana-con-ia/"&gt;Reseña de «Advancing mathematics by guiding human intuition with AI»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/28-resena-de-shaping-the-future-of-mathematics-in-the-age-of-ai/"&gt;Reseña de «Shaping the future of mathematics in the age of AI»&lt;/a&gt;. #AI4Math #LeanProver #ITP&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/29-resena-de-what-happens-when-ai-starts-checking-mathematicians-work/"&gt;Reseña de «What happens when AI starts checking mathematicians’ work»&lt;/a&gt;. #AI4Math #ITP #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/29-resena-de-as-ai-keeps-improving-mathematicians-struggle-to-foretell-their-own-future/#AI4Math"&gt;Reseña de «As AI keeps improving, mathematicians struggle to foretell their own future»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/29-resena-de-mathematics-in-the-library-of-babel/"&gt;Reseña de «Mathematics in the library of Babel»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/29-embracing-ai-and-formalization-experimenting-with-tomorrows-mathematical-tools/"&gt;Reseña de «Embracing AI and formalization: Experimenting with tomorrow’s mathematical tools»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/31-resena-de-this-startup-wants-to-change-how-mathematicians-do-math/"&gt;Reseña de «This startup wants to change how mathematicians do math»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/04/01-la-ia-acaba-de-hacerle-una-pregunta-incomoda-a-la-fisica-teorica-que-pasa-si-algunos-de-sus-resultados-mas-citados-nunca-fueron-tan-solidos-como-parecian/"&gt;Reseña de «La IA acaba de hacerle una pregunta incómoda a la física teórica&lt;/a&gt;. Qué pasa si algunos de sus resultados más citados nunca fueron tan sólidos como parecían». #LeanProver #ITP&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/04/02-short-proofs-in-combinatorics-and-number-theory/"&gt;Reseña de «Short proofs in combinatorics and number theory»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/31-mathematical-methods-and-human-thought-in-the-age-of-ai/"&gt;Reseña de «Mathematical methods and human thought in the age of AI»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/11-resena-de-mathematicians-in-the-age-of-ai/"&gt;Reseña de «Mathematicians in the age of AI»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/03/10-resena-de-ai-for-mathematical-and-scientific-discovery/"&gt;Reseña de «AI for mathematical and scientific discovery»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/02/09-accelerating-mathematics/"&gt;Reseña de «Accelerating mathematics»&lt;/a&gt;. #AI4Math #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/02/27-resena-de-ai-that-can-prove-its-right-verification-as-the-missing-layer-in-ai/"&gt;Reseña de «AI that can prove it’s right: Verification as the missing layer in AI»&lt;/a&gt;. #AI4Math #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/04/03-type-checked-compliance-deterministic-guardrails-for-agentic-financial-systems-using-lean-4-theorem-proving/"&gt;Reseña de «Type-checked compliance: Deterministic guardrails for agentic financial systems using Lean 4 theorem proving»&lt;/a&gt;. #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/04/03-what-godel-discovered/"&gt;Reseña de «What Gödel discovered»&lt;/a&gt;. #Logic #Math #Lisp #Programming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/04/03-why-lean/"&gt;Reseña de «Why Lean?»&lt;/a&gt;. #LeanProver #ITP #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/04/03-deepseek-prover-v2-7b-bridges-the-gap-between-mathematical-thought-and-formal-code/"&gt;Reseña de «DeepSeek-Prover-V2-7B bridges the gap between mathematical thought and formal code»&lt;/a&gt;. #AI4Math #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/02/03-recent-advances-in-llms-for-mathematics/"&gt;Reseña de «Recent advances in LLMs for mathematics»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/01/31-aristotle-an-ai-theorem-prover-using-lean/"&gt;Reseña de «Aristotle, an AI theorem prover using Lean»&lt;/a&gt;. #AI4Math #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2026/01/24-sobre-la-resolucion-de-problemas-de-erdos-usando-inteligencia-artificial-generativa/"&gt;Reseña de «Sobre la resolución de problemas de Erdös usando inteligencia artificial generativa»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/12/09-the-story-of-erdos-problem-1026/"&gt;Reseña de «The story of Erdős problem #1026»&lt;/a&gt;. #AI4Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/12/07-50-years-of-proof-assistants/"&gt;Reseña de «50 years of proof assistants»&lt;/a&gt;. #ITP #IsabelleHOL #LeanProver #CoqProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/11/18-teaching-real-analysis-as-a-game/"&gt;Reseña de «Teaching real analysis as a game»"&lt;/a&gt;. #LeanProver #ITP #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/11/14-how-we-achieved-an-imo-medal-one-year-before-any-other-ai-system/"&gt;Reseña de «How we achieved an IMO medal, one year before any other AI system»&lt;/a&gt;. #AI #Math #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/11/08-a-new-paradigm-for-mathematical-proof/"&gt;Reseña de «A new paradigm for mathematical proof?»&lt;/a&gt;. #Math #ITP #LeanProver #Agda&lt;/li&gt;
&lt;/ul&gt;</description><category>Agda</category><category>AI</category><category>AI4Math</category><category>AlphaProof</category><category>ATP</category><category>Autoformalization</category><category>CategoryTheory</category><category>CoqProver</category><category>FunctionalProgramming</category><category>IsabelleHOL</category><category>ITP</category><category>LambdaCalculus</category><category>LeanProver</category><category>Lisp</category><category>LLMs</category><category>Logic</category><category>LogicProgramming</category><category>Math</category><category>Physics</category><category>Programming</category><category>Prolog</category><category>Racket</category><category>RocqProver</category><category>Vampire</category><guid>https://jaalonso.github.io/vestigium/posts/2026/04/04-readings_shared_04-04-26/</guid><pubDate>Sun, 05 Apr 2026 04:00:00 GMT</pubDate></item><item><title>Readings shared November 23, 2025</title><link>https://jaalonso.github.io/vestigium/posts/2025/11/24-readings_shared_11-23-25/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;The readings shared in &lt;a href="https://bsky.app/profile/jalonso.bsky.social"&gt;Bluesky&lt;/a&gt; on 23 November 2025 are:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://arstechnica.com/ai/2025/11/deepminds-latest-an-ai-for-handling-mathematical-proofs/"&gt;DeepMind’s latest: An AI for handling mathematical proofs&lt;/a&gt;. ~ Jacek Krywko. #AI #Math #LLMs #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;li&gt;&lt;a href="https://icaps25.icaps-conference.org/program/workshops/krplan-papers/wang-abdulaziz-krplan25.pdf"&gt;Verified certification of unsolvability of temporal planning problems&lt;/a&gt;. ~ David Wang, Mohammad Abdulaziz. #ITP #IsabelleHOL&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.insmi.cnrs.fr/fr/cnrsinfo/annals-formalized-mathematics-nouvel-epijournal-formalisation"&gt;Annals of Formalized Mathematics: un nouvel épi-journal dédié à la formalisation&lt;/a&gt;. #ITP #Math #LeanProver #IsabelleHOL #Rocq #Agda&lt;/li&gt;
&lt;li&gt;&lt;a href="https://afm.episciences.org/15978"&gt;Formalization of derived categories in Lean/mathlib&lt;/a&gt;. ~ Joël Riou. #ITP #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://afm.episciences.org/15954"&gt;Formalizing zeta and L-functions in Lean&lt;/a&gt;. ~ David Loeffler, Michael Stoll. #ITP #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://afm.episciences.org/16046"&gt;A complete formalization of Fermat's Last Theorem for regular primes in Lean&lt;/a&gt;. ~ Alex Best, Christopher Birkbeck, Riccardo Brasca, Eric Rodriguez Boidi, Ruben van De Velde, Andrew Yang. #ITP #LeanProver #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://afm.episciences.org/16007"&gt;Formalising the local compactness of the adele ring&lt;/a&gt;. ~ Salvatore Mercuri. #ITP #LeanProver #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://cambium.inria.fr/blog/testing-a-priority-queue-with-Monolith/"&gt;Testing a priority queue with Monolith&lt;/a&gt;. ~ François Pottier. #OCaml #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/11/21-deepminds-latest-an-ai-for-handling-mathematical-proofs/"&gt;Reseña de «DeepMind’s latest: An AI for handling mathematical proofs»&lt;/a&gt;. #AI #Math #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;/ul&gt;</description><category>Agda</category><category>AI</category><category>AlphaProof</category><category>FunctionalProgramming</category><category>IsabelleHOL</category><category>ITP</category><category>LeanProver</category><category>LLMs</category><category>Math</category><category>OCaml</category><category>Rocq</category><guid>https://jaalonso.github.io/vestigium/posts/2025/11/24-readings_shared_11-23-25/</guid><pubDate>Mon, 24 Nov 2025 04:00:00 GMT</pubDate></item><item><title>Reseña de «DeepMind’s latest - An AI for handling mathematical proofs»</title><link>https://jaalonso.github.io/vestigium/posts/2025/11/21-deepminds-latest-an-ai-for-handling-mathematical-proofs/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;En el artículo &lt;a href="https://arstechnica.com/ai/2025/11/deepminds-latest-an-ai-for-handling-mathematical-proofs/"&gt;«DeepMind’s latest: An AI for handling mathematical proofs»&lt;/a&gt;, se presenta a AlphaProof, un sistema de inteligencia artificial capaz de razonar y realizar demostraciones matemáticas complejas. En la Olimpiada Internacional de Matemáticas de 2024, el sistema alcanzó el nivel de un medallista de plata, quedándose a un solo punto de obtener el oro.&lt;/p&gt;
&lt;h2&gt;El desafío del razonamiento lógico&lt;/h2&gt;
&lt;p&gt;Históricamente, los ordenadores han sido excelentes calculando, pero mediocres "entendiendo" la lógica necesaria para las demostraciones matemáticas. Los modelos de lenguaje actuales (como ChatGPT) funcionan mediante estadística y predicción de palabras, lo que a menudo resulta en respuestas que "suenan" bien pero son matemáticamente incorrectas. Para solucionar esto, DeepMind necesitaba un sistema que garantizase certeza absoluta.&lt;/p&gt;
&lt;h2&gt;Cómo funciona: La combinación de tres elementos&lt;/h2&gt;
&lt;p&gt;Para lograr este hito, AlphaProof combinó varias estrategias:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Lenguaje formal (Lean):&lt;/strong&gt; Utilizaron Lean, un software que permite escribir y verificar demostraciones matemáticas. Como había pocos datos de entrenamiento en este formato, usaron una versión de Gemini para traducir millones de problemas de lenguaje natural a Lean.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Aprendizaje por refuerzo (Estilo AlphaZero):&lt;/strong&gt; Emplearon una arquitectura similar a la que domina el ajedrez o el Go. Una red neuronal aprende mediante prueba y error, recompensando las demostraciones correctas y elegantes, combinada con un algoritmo de búsqueda en árbol para explorar posibles pasos lógicos.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Aprendizaje en tiempo de prueba (TTRL):&lt;/strong&gt; Esta es la gran innovación. Ante un problema muy difícil, la IA genera variaciones del mismo (algunas más simples, otras más generales) para "practicar" y aprender sobre la marcha, emulando cómo un humano intenta resolver versiones simplificadas de un problema antes de atacar el original.&lt;/p&gt;
&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;Resultados&lt;/h2&gt;
&lt;p&gt;AlphaProof logró resolver el problema más difícil de la competición (el sexto), algo que solo consiguieron 6 de los 609 participantes humanos. En conjunto con AlphaGeometry 2 (que se encargó de un problema de geometría para el que AlphaProof no estaba optimizado), el sistema sumó 28 puntos.&lt;/p&gt;
&lt;h2&gt;Las limitaciones: Tiempo y coste&lt;/h2&gt;
&lt;p&gt;A pesar del éxito, el sistema tiene desventajas significativas frente a los humanos:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Recursos desproporcionados:&lt;/strong&gt; Mientras que los estudiantes tienen 4,5 horas por sesión, AlphaProof tardó días en resolver los problemas, utilizando una inmensa potencia de cálculo (cientos de días de procesamiento TPU).&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Coste prohibitivo:&lt;/strong&gt; Actualmente, el coste de ejecutar este sistema es inviable para la mayoría de los investigadores.&lt;/p&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;p&gt;&lt;strong&gt;Dependencia humana:&lt;/strong&gt; Los problemas tuvieron que ser traducidos manualmente a Lean antes de que la IA pudiera procesarlos.&lt;/p&gt;
&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;El futuro&lt;/h2&gt;
&lt;p&gt;El objetivo de DeepMind no es ganar concursos como la IMO, sino optimizar esta tecnología para que sea menos costosa y pueda contribuir a la investigación matemática profesional, ayudando a descubrir nuevos conceptos en lugar de solo resolver los ya conocidos.&lt;/p&gt;</description><category>AI</category><category>AlphaProof</category><category>ITP</category><category>LeanProver</category><category>Math</category><guid>https://jaalonso.github.io/vestigium/posts/2025/11/21-deepminds-latest-an-ai-for-handling-mathematical-proofs/</guid><pubDate>Fri, 21 Nov 2025 09:30:00 GMT</pubDate></item><item><title>Readings shared November 14, 2025</title><link>https://jaalonso.github.io/vestigium/posts/2025/11/15-readings_shared_11-14-25/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;The readings shared in &lt;a href="https://bsky.app/profile/jalonso.bsky.social"&gt;Bluesky&lt;/a&gt; on 14 November 2025 are:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://youtu.be/VI63OwWMxJM"&gt;An introduction to formal real analysis (Lecture 18: Rearrangements)&lt;/a&gt;. ~ Alex Kontorovich. #ITP #LeanProver #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/choice-trees-representing-and-reasoning-about-nondeterministic-recursive-and-impure-programs-in-rocq/797B95A45813CAE54CFFFA3D7B163BA4"&gt;Choice trees: Representing and reasoning about nondeterministic, recursive, and impure programs in Rocq&lt;/a&gt;. ~ Nicolas Chappe et als. #ITP #Rocq&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/practical-formalization-of-monadic-equational-reasoning-in-dependenttype-theory/B59B87DE000F48B9807F24AEDB11452E"&gt;A practical formalization of monadic equational reasoning in dependent-type theory&lt;/a&gt;. ~ Reynald Affeldt, Jacques Garrigue, Takafumi Saikawa. #ITP #CoqProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/towards-typedirected-compiler-calculation/247DC1BC3BB40F8AA99DF685861CDF8A"&gt;Towards type-directed compiler calculation&lt;/a&gt;. ~ Wouter Swierstra. #ITP #Agda&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/binary-searchthink-positive/1F974DF9C6475D19B32E18C93A2DCC1C"&gt;Binary search—think positive&lt;/a&gt;. ~ Alexander Dinges, Ralf Hinze. #ITP #Agda&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/checking-equivalence-in-a-nonstrict-language/82D1E4CE7541BDDF5B954B471549C44"&gt;Checking equivalence in a non-strict language&lt;/a&gt;. ~ John Charles Kolesar, Ruzica Piskac, William Triest Hallahan.7#supplementary-materials #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/automatically-testing-console-io-behavior-of-student-submissions-in-haskell/20BF7DCA6330A2115C2C2B9BA47AB2E0"&gt;Automatically testing console I/O behavior of student submissions in Haskell&lt;/a&gt;. ~ Oliver Westphal, Janis Voigtländer. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/parallel-dualnumbers-reverse-ad/ECFEF5DE72D5CB7C5BBA0AD7C203BF38"&gt;Parallel dual-numbers reverse AD&lt;/a&gt;. ~ Tom J. Smeding, Matthijs I.L. Vákár. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/domainspecific-tensor-languages/19B95794B66C66E117DFCFC7A21E22D5"&gt;Domain-specific tensor languages&lt;/a&gt;. ~ Jean-Philippe Bernardy, Patrik Jansson. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/turner-bird-eratosthenes-an-eternal-burning-thread/32E2EDF5D5EAEC95F13D313BC97B86F0"&gt;Turner, Bird, Eratosthenes: An eternal burning thread&lt;/a&gt;. ~ Jeremy Gibbons. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/how-much-is-in-a-square-calculating-functional-programs-with-squares/F48258008F47DC9F53AA2E61B4E511A7"&gt;How much is in a square? Calculating functional programs with squares&lt;/a&gt;. ~ Jose Nuno Oliveira. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/experiences-of-early-assessment-to-teach-functional-programming/667F18D3611D4A3A7C7C6A2A2FA4FB3C"&gt;Experiences of early assessment to teach functional programming&lt;/a&gt;. ~ Peter Chapman. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/you-could-have-invented-fenwick-trees/B4628279D4E54229CED97249E96F721D"&gt;You could have invented Fenwick trees&lt;/a&gt;. ~ Brent Yorgey. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.mlabs.city/blog/patterns-and-paradoxes"&gt;Patterns and paradoxes: The logic of pattern synonyms&lt;/a&gt;. ~ Koz Ross. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.youtube.com/live/EcV4rb-Czfg"&gt;The Haskell Unfolder Episode 49: Shrinking&lt;/a&gt;. ~ Edsko de Vries, Andres Löh. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.youtube.com/live/-zxxl-WuwuE"&gt;The Haskell Unfolder Episode 50: Singletons&lt;/a&gt;. ~ Edsko de Vries, Andres Löh. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.youtube.com/live/3XeIoA0aVLQ"&gt;The Haskell Unfolder Episode 51: Typed servers using sop-core&lt;/a&gt;. ~ Edsko de Vries, Andres Löh. #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cambridge.org/core/journals/journal-of-functional-programming/article/tail-recursion-modulo-context-an-equational-approach-extended-version/CCFFA4B8583A4924F002B0C377394303"&gt;Tail recursion modulo context: An equational approach&lt;/a&gt;. ~ Daan Leijen, Anton Felix Lorenzen. #OCaml #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.tomzahavy.com/post/how-we-achieved-an-imo-medal-one-year-before-everyone-else"&gt;How we achieved an IMO medal, one year before any other AI system&lt;/a&gt;. ~ Tom Zahavy. #AI #Math #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.nature.com/articles/s41586-025-09833-y"&gt;Olympiad-level formal mathematical reasoning with reinforcement learning&lt;/a&gt;. ~ Thomas Hubert et als. #AI #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/11/14-olympiad-level-formal-mathematical-reasoning-with-reinforcement-learning/"&gt;Reseña de «Olympiad-level formal mathematical reasoning with reinforcement learning»&lt;/a&gt;. #AI #Math #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/exercitium/posts/2014/11/21-cuantificadores_sobre_listas/"&gt;#Exercitium: Cuantificadores sobre listas&lt;/a&gt;. #Haskell #ProgramaciónFuncional&lt;/li&gt;
&lt;/ul&gt;</description><category>Agda</category><category>AI</category><category>AlphaProof</category><category>CoqProver</category><category>FunctionalProgramming</category><category>Haskell</category><category>ITP</category><category>LeanProver</category><category>Math</category><category>OCaml</category><category>Rocq</category><guid>https://jaalonso.github.io/vestigium/posts/2025/11/15-readings_shared_11-14-25/</guid><pubDate>Sat, 15 Nov 2025 04:00:00 GMT</pubDate></item><item><title>Reseña de «How we achieved an IMO medal, one year before any other AI system»</title><link>https://jaalonso.github.io/vestigium/posts/2025/11/14-how-we-achieved-an-imo-medal-one-year-before-any-other-ai-system/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;En el artículo &lt;a href="https://www.tomzahavy.com/post/how-we-achieved-an-imo-medal-one-year-before-everyone-else"&gt;«How we achieved an IMO medal, one year before any other AI system»&lt;/a&gt; se explica cómo AlphaProof logró un nivel de medalla de plata en la Olimpiada Internacional de Matemáticas. Este hito, alcanzado un año antes que cualquier otra IA, superó un gran desafío para la inteligencia artificial en una competición de élite para jóvenes matemáticos.&lt;/p&gt;
&lt;p&gt;El avance crucial fue el "Aprendizaje por Refuerzo en Tiempo de Prueba" (TTRL). Esta técnica permitía al agente crear variaciones de los problemas y entrenar con ellas durante la propia competencia. Aunque consumía muchos recursos y tuvo resultados iniciales modestos, su refinamiento fue clave para lograr tres demostraciones completas tras tres días de prueba.&lt;/p&gt;
&lt;p&gt;El autor señala futuros desafíos como la dependencia del demostrador Lean y la limitada cantidad de problemas matemáticos únicos. Su visión es desarrollar agentes que no solo resuelvan problemas, sino que generen sus propias preguntas y construyan teorías novedosas de forma autónoma.&lt;/p&gt;</description><category>AI</category><category>AlphaProof</category><category>ITP</category><category>LeanProver</category><category>Math</category><guid>https://jaalonso.github.io/vestigium/posts/2025/11/14-how-we-achieved-an-imo-medal-one-year-before-any-other-ai-system/</guid><pubDate>Fri, 14 Nov 2025 12:00:00 GMT</pubDate></item><item><title>Reseña de «Olympiad-level formal mathematical reasoning with reinforcement learning»</title><link>https://jaalonso.github.io/vestigium/posts/2025/11/14-olympiad-level-formal-mathematical-reasoning-with-reinforcement-learning/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;En el artículo &lt;a href="https://www.nature.com/articles/s41586-025-09833-y"&gt;«Olympiad-level formal mathematical reasoning with reinforcement learning»&lt;/a&gt; se presenta &lt;a href="https://www.julian.ac/blog/2025/11/13/alphaproof-paper/"&gt;AlphaProof&lt;/a&gt;, un agente de IA que utiliza &lt;a href="https://es.wikipedia.org/wiki/Aprendizaje_por_refuerzo"&gt;aprendizaje por refuerzo&lt;/a&gt; para resolver problemas matemáticos complejos en el asistente de pruebas &lt;a href="https://en.wikipedia.org/wiki/Lean_(proof_assistant)"&gt;Lean&lt;/a&gt;. Combina una red neuronal con búsqueda en árbol para encontrar demostraciones formalmente verificables.&lt;/p&gt;
&lt;p&gt;El sistema fue entrenado con millones de problemas auto-formalizados y emplea una innovadora adaptación en tiempo de prueba para los problemas más difíciles. Demostró capacidades excepcionales resolviendo problemas de competiciones como la &lt;a href="https://www.imo-official.org"&gt;Olimpiada Internacional de Matemáticas&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;En la &lt;a href="https://www.imo2024.uk/"&gt;IMO 2024&lt;/a&gt;, AlphaProof resolvió tres problemas no geométricos, incluyendo el más difícil. Combinado con &lt;a href="https://en.wikipedia.org/wiki/AlphaGeometry#AlphaGeometry_2"&gt;AlphaGeometry 2&lt;/a&gt;, el sistema logró una puntuación equivalente a &lt;a href="https://deepmind.google/blog/ai-solves-imo-problems-at-silver-medal-level/"&gt;medalla de plata&lt;/a&gt;, marcando un hito en el razonamiento matemático automatizado.&lt;/p&gt;</description><category>AI</category><category>AlphaProof</category><category>ITP</category><category>LeanProver</category><category>Math</category><guid>https://jaalonso.github.io/vestigium/posts/2025/11/14-olympiad-level-formal-mathematical-reasoning-with-reinforcement-learning/</guid><pubDate>Fri, 14 Nov 2025 09:00:00 GMT</pubDate></item><item><title>Readings shared October 03, 2025</title><link>https://jaalonso.github.io/vestigium/posts/2025/10/04-readings_shared_10-03-25/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;The readings shared in &lt;a href="https://bsky.app/profile/jalonso.bsky.social"&gt;Bluesky&lt;/a&gt; on 03 October 2025 are:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.renaissancephilanthropy.org/news-and-insights/kevin-buzzard-and-alex-kontorovich-on-the-future-of-formal-mathematics-a-mathlib-initiative-interview"&gt;Kevin Buzzard and Alex Kontorovich on the future of formal mathematics: A Mathlib initiative interview&lt;/a&gt;. ~ Oliver Nash. #ITP #LeanProver #Mathlib #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://mathlib-initiative.org/"&gt;The Mathlib initiative (Building the digital foundation of mathematics)&lt;/a&gt;. #ITP #LeanProver #Mathlib #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://youtu.be/STTvEmBM3rg"&gt;Mathematics in the age of automated proofs: What do we tell our students about AI? ~ Akshay Venkatesh&lt;/a&gt;. #Math #AI&lt;/li&gt;
&lt;li&gt;&lt;a href="https://youtu.be/uhwfTOUIeiw"&gt;Mathematics in the age of automated proofs: AlphaProof (From the lab into your hands)&lt;/a&gt;. ~ Thomas Hubert. #AI #Math #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;li&gt;&lt;a href="https://youtu.be/p-Quz60JMKg"&gt;Mathematics in the age of automated proofs: After Math (historical perspectives on automated intelligence)&lt;/a&gt;. ~ Stephanie Dick. #AI #Math #ATP&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/exercitium/posts/2014/07/17-insercion_en_arboles_binarios_de_busqueda/"&gt;#Exercitium: Inserción en árboles binarios de búsqueda&lt;/a&gt;. #Haskell #ProgramaciónFuncional #Matemáticas&lt;/li&gt;
&lt;/ul&gt;</description><category>AI</category><category>AlphaProof</category><category>ATP</category><category>FunctionalProgramming</category><category>Haskell</category><category>ITP</category><category>LeanProver</category><category>Math</category><category>Mathlib</category><guid>https://jaalonso.github.io/vestigium/posts/2025/10/04-readings_shared_10-03-25/</guid><pubDate>Sat, 04 Oct 2025 04:00:00 GMT</pubDate></item><item><title>Readings shared June 10, 2025</title><link>https://jaalonso.github.io/vestigium/posts/2025/06/11-readings_shared_06-10-25/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;The readings shared in &lt;a href="https://bsky.app/profile/jalonso.bsky.social"&gt;Bluesky&lt;/a&gt; on 10 June 2025 are&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://lawrencecpaulson.github.io//2025/06/09/Inductive_Definitions.html"&gt;Inductive definitions&lt;/a&gt;. ~ Lawrence Paulson. #ITP #IsabelleHOL #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://terrytao.wordpress.com/wp-content/uploads/2025/06/math-experiments.pdf"&gt;The equational theories project: advancing collaborative mathematical research at scale&lt;/a&gt;. ~ Terence Tao et als. #ITP #LeanProver #Math #AIforMath&lt;/li&gt;
&lt;li&gt;&lt;a href="https://mathscholar.org/2025/06/new-ai-stuns-mathematicians-with-its-problem-solving-skill/"&gt;New AI stuns mathematicians with its problem-solving skill&lt;/a&gt;. ~ David H Bailey. #AI #LLMs #Math #AIforMath #ITP #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2506.07477"&gt;Premise selection for a Lean hammer&lt;/a&gt;. ~ Thomas Zhu, Joshua Clune, Jeremy Avigad, Albert Qiaochu Jiang, Sean Welleck. #ITP #LeanProver #AI&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2506.04592v1"&gt;Safe: Enhancing mathematical reasoning in large language models via retrospective step-aware formal verification&lt;/a&gt;. ~ Chengwu Liu et als. #LLMs #Math #ITP #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.youtube.com/live/xZIqn4V6O0A"&gt;AlphaProof: When RL meets formal maths&lt;/a&gt;. ~ Thomas Hubert. #AI #Math #AIforMath #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.youtube.com/live/WdXnMrT1X_k"&gt;AI for Math: The future of collaborative discovery&lt;/a&gt;. ~ Mateja Jamnik. #AIforMath&lt;/li&gt;
&lt;li&gt;&lt;a href="https://arxiv.org/abs/2506.04575v1"&gt;Are LLMs reliable translators of logical reasoning across lexically diversified contexts?&lt;/a&gt; ~ Qingchuan Li et als. #LLMs #Math #ATP #Prover9&lt;/li&gt;
&lt;li&gt;&lt;a href="https://coyotetracks.org/blog/dr-neckbeard-emacs/"&gt;Dr. Neckbeard, or how I learned to stop worrying and love Emacs&lt;/a&gt;. ~ Watts Martin. #Emacs&lt;/li&gt;
&lt;li&gt;&lt;a href="https://bib.us.es/sites/bib3.us.es/files/investiga52.pdf"&gt;Revistas, editoriales y congresos depredadores&lt;/a&gt;. #Investigación&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/06/10-alphaproof-aprendizaje-por-refuerzo-aplicado-a-la-demostracion-matematica/"&gt;AlphaProof: Aprendizaje por refuerzo aplicado a la demostración matemática&lt;/a&gt;. #AI #Math #AIforMath #ITP #LeanProver #AlphaProof&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/06/10-el-futuro-de-las-matematicas-descubrimiento-colaborativo-entre-humanos-y-maquinas/"&gt;El futuro de las matemáticas: Descubrimiento colaborativo entre humanos y máquinas&lt;/a&gt;. #AIforMath #AI #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/vestigium/posts/2025/06/10-el-proyecto-etp-un-caso-de-estudio-en-investigacion-matematica-colaborativa-y-formalizada/"&gt;El proyecto ETP (Un caso de estudio en investigación matemática colaborativa y formalizada)&lt;/a&gt;. #AIforMath #ITP #LeanProver #Math&lt;/li&gt;
&lt;/ul&gt;</description><category>AI</category><category>AIforMath</category><category>AlphaProof</category><category>ATP</category><category>Emacs</category><category>IsabelleHOL</category><category>ITP</category><category>LeanProver</category><category>LLMs</category><category>Math</category><category>Prover9</category><guid>https://jaalonso.github.io/vestigium/posts/2025/06/11-readings_shared_06-10-25/</guid><pubDate>Wed, 11 Jun 2025 09:06:00 GMT</pubDate></item><item><title>Reseña de «AlphaProof: When RL meets formal maths»</title><link>https://jaalonso.github.io/vestigium/posts/2025/06/10-alphaproof-when-rl-meets-formal-maths/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;En su conferencia de ayer &lt;a href="https://www.youtube.com/live/xZIqn4V6O0A"&gt;«AlphaProof: When RL meets formal maths»&lt;/a&gt;, Thomas Hubert de Google DeepMind presentó &lt;a href="https://deepmind.google/discover/blog/ai-solves-imo-problems-at-silver-medal-level/"&gt;AlphaProof&lt;/a&gt;, un sistema de inteligencia artificial que aplica los principios del &lt;a href="https://en.wikipedia.org/wiki/Reinforcement_learning"&gt;aprendizaje por refuerzo&lt;/a&gt; (RL, del inglés "&lt;em&gt;Reinforcement learning&lt;/em&gt;") al dominio de las matemáticas formales. El concepto fundamental radica en que los asistentes de demostración como &lt;a href="https://en.wikipedia.org/wiki/Lean_(proof_assistant)"&gt;Lean&lt;/a&gt; proporcionan el entorno perfecto para el RL: un espacio de experimentación masiva con retroalimentación inequívoca (una prueba matemática es correcta o incorrecta, sin ambigüedades). Siguiendo el paradigma exitoso de sistemas como AlphaGo Zero, AlphaProof está diseñado para generar conocimiento matemático de forma autónoma, trascendiendo la mera imitación de demostraciones humanas preexistentes.&lt;/p&gt;
&lt;p&gt;La arquitectura de AlphaProof se estructura en un proceso de múltiples fases que combina diferentes técnicas de aprendizaje automático. Inicialmente, emplea modelos de lenguaje para autoformalizar problemas matemáticos expresados en lenguaje natural hacia el código formal de Lean, generando así un extenso conjunto de datos de entrenamiento. Su modelo demostrador experimenta dos etapas: primero, un entrenamiento supervisado utilizando la biblioteca &lt;a href="https://leanprover-community.github.io/mathlib-overview.html"&gt;mathlib&lt;/a&gt;, seguido de un refinamiento intensivo mediante RL que resuelve millones de problemas matemáticos. Para desafíos particularmente complejos —como los de la Olimpiada Internacional de Matemáticas (IMO)—, el sistema implementa una estrategia de adaptación en tiempo real, generando y resolviendo múltiples variantes de un problema para desarrollar gradualmente la intuición necesaria.&lt;/p&gt;
&lt;p&gt;Los resultados validan el potencial transformador de AlphaProof como herramienta matemática colaborativa. El sistema alcanzó una puntuación equivalente a una &lt;a href="https://deepmind.google/discover/blog/ai-solves-imo-problems-at-silver-medal-level/"&gt;medalla de plata en la IMO&lt;/a&gt; y, durante una demostración en vivo durante la conferencia, asistió exitosamente a un matemático a completar los pasos de una prueba compleja relacionada con la función zeta de Riemann. Hubert enfatiza que el objetivo trasciende las competiciones académicas: la meta fundamental es contribuir significativamente a la investigación matemática contemporánea, convirtiendo AlphaProof en una útil para la comunidad matemática que facilite el descubrimiento y verificación de nuevas verdades matemáticas.&lt;/p&gt;</description><category>AI</category><category>AlphaProof</category><category>ITP</category><category>LeanProver</category><category>Math</category><category>RL</category><guid>https://jaalonso.github.io/vestigium/posts/2025/06/10-alphaproof-when-rl-meets-formal-maths/</guid><pubDate>Tue, 10 Jun 2025 11:50:00 GMT</pubDate></item><item><title>Readings shared April 3, 2025</title><link>https://jaalonso.github.io/vestigium/posts/2025/04/03-readings_shared_04-03-25/</link><dc:creator>José A. Alonso</dc:creator><description>&lt;p&gt;The readings shared in &lt;a href="https://bsky.app/profile/jalonso.bsky.social"&gt;Bluesky&lt;/a&gt; on 3 April 2025 are&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.isa-afp.org/entries/Morley_Theorem.html"&gt;Morley's theorem (in Isabelle/HOL)&lt;/a&gt;. ~ Benjamin Puyobro. #ITP #IsabelleHOL #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://books.google.com/books?id=4Xc2EQAAQBAJ"&gt;Category theory using Haskell (An introduction with Moggi and Yoneda)&lt;/a&gt;. ~ Shuichi Yukita. #CategoryTheory #Math #Haskell #FunctionalProgramming&lt;/li&gt;
&lt;li&gt;&lt;a href="https://repository.tudelft.nl/file/File_a98b6c93-4017-42c5-bb8f-df68da0d7034"&gt;LeanSolver: Solving theorems through large language models and search&lt;/a&gt;. ~ Avi Luciano Halevy  #LLMs #ITP #LeanProver&lt;/li&gt;
&lt;li&gt;&lt;a href="https://youtu.be/TFBzP78Jp6A"&gt;AlphaProof: when reinforcement learning meets formal mathematics&lt;/a&gt;. ~ Thomas Hubert. #AlphaProof #AI #Math&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/exercitium/posts/2014/05/23-biparticiones_de_una_lista/"&gt;#Exercitium: Biparticiones de una lista&lt;/a&gt;. #Haskell #ProgramaciónFuncional&lt;/li&gt;
&lt;li&gt;&lt;a href="https://jaalonso.github.io/calculemus/posts/2021/06/04-union_con_interseccion_general/"&gt;Demostraciones con Lean4 y con Isabelle/HOL de "s ∪ (⋂ i, A i) = ⋂ i, (A i ∪ s)"&lt;/a&gt;. #LeanProver #IsabelleHOL #Math #Calculemus&lt;/li&gt;
&lt;/ul&gt;</description><category>AI</category><category>AlphaProof</category><category>CategoryTheory</category><category>FunctionalProgramming</category><category>Haskell</category><category>IsabelleHOL</category><category>ITP</category><category>LeanProver</category><category>LLMs</category><category>Math</category><guid>https://jaalonso.github.io/vestigium/posts/2025/04/03-readings_shared_04-03-25/</guid><pubDate>Thu, 03 Apr 2025 04:00:00 GMT</pubDate></item></channel></rss>