I'm thrilled to share that I have successfully defended my PhD thesis. The techniques proposed in this thesis were evaluated on APIs from major platforms, leading to bug fixes and documentation updates in services like GitHub and Amadeus Hotel. (1/3)

And the Best #BPM2025 Dissertation Award goes to @antoniommrr!! 🎉🎉 It couldn't be otherwise. So proud of our child!! cc: @prof_JGEnriquez

@gbavota @USI_INF A big thank you to my supervisors @sergio_segura2 and @TonioRuizCortes, as well as to @gbavota and @a_martinlopez, for their invaluable guidance and collaboration throughout this journey. Looking forward to travelling to South Korea to present this work at ASE 2025✈️

@gbavota @USI_INF We also introduce OKAMI, a dataset of 10,645 manually annotated test oracles across 17 operations, designed as a benchmark for future research. 👉

@gbavota @USI_INF In our evaluation on 12 industrial APIs, SATORI generated hundreds of valid oracles per operation, revealed 18 real bugs in widely used APIs, and proved complementary to AGORA+, my earlier dynamic oracle generation technique.

@gbavota @USI_INF SATORI (Static API Test ORacle Inference) is a static, black-box approach that leverages LLMs to automatically generate test oracles for REST APIs by analyzing only their OpenAPI specifications—no prior execution required.

This work marks the final contribution of my PhD, and it is the result of a collaboration with Prof. @gbavota during my research stay at Università della Svizzera italiana, @USI_INF.

Thrilled to share that our paper SATORI: Static Test Oracle Generation for REST APIshas been accepted at the 40th IEEE/ACM International Conference on Automated Software Engineering (ASE 2025) 🎉 👉 Pre-print:

Muchos amigos tuiteros pidieron ayer recomendaciones sobre plumas estilográficas de aspecto clásico que sean bonitas, baratas y eficaces. Éstas son dos de mis favoritas entre ellas: la Hongdian N12 (36 €) y la icónica Lamy Safari (23 €).

A sincere thank you as well to the members of the Thesis Examination Committee (Aitor Arrieta, Alessandra Gorla, Carlos Müller, Alberto Martín, and Ana Belén Sánchez) for their thoughtful questions and constructive feedback during the defense. Excited for what comes next! (3/3)

I am deeply grateful to my supervisors, Professors Sergio Segura and Antonio Ruiz, for their unwavering support, insightful guidance, and continuous encouragement throughout every stage of this journey. (2/3)

🔴 ¡NOTEBOOKLM CON VIDEO! Os lo he contado en el directo: NotebookLM próximamente no sólo te generará podcast de tus documentos y fuentes, sino también piezas divulgativas animadas con gráficos, diagramas, fotos y en un futuro no muy lejano esperaría incluso clips de vídeos de

I still think about this scene a lot. Kazuhiro Nakaya killed it is as Ichiban.

Mondragon team in action yesterday and today at @ICSEconf

goodbye, GPT-4. you kicked off a revolution. we will proudly keep your weights on a special hard drive to give to some historians in the future.

Elon’s nonstop actions against us are just bad-faith tactics to slow down OpenAI and seize control of the leading AI innovations for his personal benefit. Today, we counter-sued to stop him.

We’re releasing PaperBench, a benchmark evaluating the ability of AI agents to replicate state-of-the-art AI research, as part of our Preparedness Framework. Agents must replicate top ICML 2024 papers, including understanding the paper, writing code, and executing experiments.

I watched ”MICKEY 17” in IMAX. It’s an unmistakably Bong Joon-Ho film, packed with his DNA and signature “memes,” yet incapable of mass production (or duplication). While a big-budget film aimed at a global audience, it brims with the auteur-driven sensibilities and disruptive

Some people today are discouraging others from learning programming on the grounds AI will automate it. This advice will be seen as some of the worst career advice ever given. I disagree with the Turing Award and Nobel prize winner who wrote, “It is far more likely that the

Detecting misbehavior in frontier reasoning models Chain-of-thought (CoT) reasoning models “think” in natural language understandable by humans. Monitoring their “thinking” has allowed us to detect misbehavior such as subverting tests in coding tasks, deceiving users, or giving