Jose Manuel de la Chica

Global Head of Santander AI Lab | Leading frontier AI with responsibility. Shaping the future with clarity and purpose.

[Jose Manuel de la Chica @Delachica]

Si la ciencia hecha con IA no se hace bien, las empresas minarán su propia credibilidad a muchos niveles. La ciencia funciona porque tiene sus propias reglas y si no se siguen, entraremos con mal pie en una etapa de progreso que debiera seguir un fair play con siglos de historia.

[Valerio Capraro @ValerioCapraro]

BREAKING: OpenAI might have stolen another major proof.

In a detailed Mastodon post, which I report in full in the comments, Andreas Thom presents several pieces of evidence suggesting that OpenAI may have trained Astra on conversations in which he and Gábor Kun were working on Gromov’s soficity conjecture, one of the ten problems OpenAI later announced Astra had solved.

I know Andreas. We met several times early in our careers. He is an exceptional mathematician, a leading expert on sofic and hyperlinear groups, and one of the most respected scholars in the field. He has spent two decades working on this problem.

If his account is correct, this is not a minor dispute over attribution. It would mean that unpublished human work was absorbed into a model and then presented to the world as a breakthrough by the model itself.

And if the allegations raised by Levent Alpöge, Tristan Buckmaster, and now Andreas Thom are all substantiated, we are no longer looking at isolated incidents.

We may be looking at one of the greatest intellectual scandals in the history of science.

AI is not discovering new mathematics.

AI is stealing human discovery.

[Jose Manuel de la Chica @Delachica]

Nos hace falta como nunca bien periodismo de Tecnología y Ciencia. Noticias como está son siempre positivas para todo el sector tecnológico en un momento de cambios apasionantes. Enhorabuena @michaelMCsaez

[Michael McLoughlin 🐦 @michaelMCsaez]

Personal news. Después de 9 años en El Confidencial, la pasada semana me propusieron ponerme al frente de la sección de Tecnología y Ciencia. Y, bueno, aquí estamos, con muchas ganas y algo de vértigo. Así que ya saben donde estoy para cualquier pista o historia.

[Jose Manuel de la Chica @Delachica]

Tres pistas buenísimas a seguir para distinguir entre señal y ruido acerca de los últimos acontecimientos alrededor de la IA (AGI, Astra, incidentes varios, etc). 👇

[Manuel Ángel Méndez @m_angelmendez]

Hemos intentado explicar en un vídeo la distopía de las últimas semanas en la industria de la IA. Que si civilizaciones artificiales, que si AGI... Hablamos con tres de los mejores expertos en ciber e IA: @ZackKorman @patowc y @cfenollosa. Algunas pistas:

[Jose Manuel de la Chica @Delachica]

La "aleatoriedad" es la mayoría de las veces solo desconocimiento de un conjunto de reglas.

[MindArch @mindarchx]

https://t.co/RcrOCT3O0g

[Jose Manuel de la Chica @Delachica]

A veces, la mejor manera de ganar una guerra, es no entrar en ella.

[Javier Martin @liderarmente]

Lo de la bolsa es flipante.

Apple que se supone que había perdido la carrera de la IA +60% en un año.

Microsoft con su Copilot y la participación en OpenAI -23% en el mismo periodo.

[Jose Manuel de la Chica @Delachica]

@adriano_galano @bancosantander @FGCSIC Gracias @adriano_galano. Viniendo de alguien como tú, estas palabras significan mucho. Lo intentamos. Estamos viviendo un momento apasionante y hay demasiadas cosas interesantes que hacer y aprender para perder el tiempo en ruido y humo como bien dices ;) un abrazo!!

[Jose Manuel de la Chica @Delachica]

Ya está disponible «La nueva frontera europea de la IA B2B», informe de @bancosantander AI Lab y @FGCSIC sobre especialización vertical de la IA, IA soberana, confianza regulatoria, tendencias, ecosistema, etc.
Lee el resumen y descargalo en:
https://t.co/8A8ri92NLR

[Jose Manuel de la Chica @Delachica]

@borjaperfra @bancosantander @FGCSIC Gracias por el feedback @borjaperfra :) estamos en ello! tienes un resumen y enlaces al doc aquí también: https://t.co/pcwLIYiogG

[Jose Manuel de la Chica @Delachica]

El vibe thinking es más interesante como actividad que el vibe coding.
El vibe coding automatiza las manos. El vibe thinking afila lo que hace la automatización.

[Andrej Karpathy @karpathy]

One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 minutes, total mess, anything goes, full stream of consciousness. Sometimes I declare it up top, something like "switching to speech recognition sorry for any typos...". Sometimes I turn it into a small interview of a few turns. But I find that the LLMs are somehow very good at reconstructing long incoherent rambles and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with. The result is that you improve the mind meld and have to correct things less from that point on.

[Jose Manuel de la Chica @Delachica]

https://t.co/64LtQSpaxy

[Jose Manuel de la Chica @Delachica]

Acabo de publicar un post largo en mi Substack: "Del Mythos al espacio latente: la parte de la IA que no estamos mirando, y por qué ya es la que más importa".
Sobre el J-lens, Mythos y lo que implica para las empresas que van a IA-first 👇

[Jose Manuel de la Chica @Delachica]

Todo por hacer en IA de frontera en los próximos años. Talento de talla mundial el que hay en @CSIC y una ocasión única para trabajar a su lado desde el Santander AI Lab gracias a este acuerdo que nos ilusiona mucho.

[CSIC @CSIC]

🏛️ #CSIC y el Santander acuerdan impulsar la investigación en IA avanzada y responsable para la banca del futuro

🖥️ El acuerdo establece el desarrollo de proyectos de investigación e innovación en IA y su conexión con nuevas capacidades computacionales

➡️https://t.co/826hWDX18h

[Jose Manuel de la Chica @Delachica]

Proud of everything we're doing at the @bancosantander AI Lab. #opensource #AI

[Simon Taylor @sytaylor]

I love this! Santander has open-sourced its open-source AI initiatives.

The bank pushed 11 repos, live this week under Apache-2.0 on the code, but the data synthetic or anonymised only.

Quite a moment for a bank this size, putting its AI control layer on the open internet for anyone to fork. This is the bit every bank has to get right.

So what is it?

→ autoguardrails: a scaffold for stress-testing LLM guardrails, jailbreaks included (can we use this LLM?)

→ "mechanical governance" for high-stakes LLM decisions, with hard gates and governance metrics (can we trust an LLM with this decision?)

→ mutatis-mutandis: discrimination testing with counterfactual comparators, straight out of a published paper (very important if you're lending!)

→ stressed-datasets: public benchmarks republished in "stressed" form to probe model robustness in that scenario

→ gen-fraud-graph: a synthetic fraud-graph generator to benchmark fraud detection (really, really cool, need to dig into this one)

→ llm_bridge: a vendor-neutral client for OpenAI, Bedrock and Gemini, so you skip the lock-in (again, how many companies are struggling with this?)

→ ralph: their own spin on the Ralph loop, the run-an-agent-in-a-loop trick from the indie AI crowd

I think I need to write a whole Rant on each of these pieces.

The most important thing for a big regulated actor is "Can you show a decision was safe, fair, auditable, and the same tomorrow as it was today." Santander published its working answer and handed it to everyone, competitors included.

Why give it away?

1. Attract talent - this is a huge signal they've got their AI act together

2. Signal internally - We have these tools, use them

3. Give regulators confidence - Here's how we work, you can audit it

(The board that signs off on releases includes Legal and the CISO. That tells you how seriously they treat it.)

I've watched banks spend years trying to govern AI behind closed doors and ship nothing. Doing it in the open, with a contributor agreement and a proper open-source office, is a faster route to getting it right.

The banks that pull ahead from here will be the ones who can prove their AI works.

@bancosantander just open-sourced a head start.

Repo is here. 👇

https://t.co/IilShwzvl2

[Jorge G. Casanova @_casanovajg]

Hace años oí a @Delachica hablar de tecnologías exponenciales y su diferencia con las habilitadoras.
Habló también de la IA como habilitadora y exponencial a la vez.
Hoy veo muchos casos de uso a mi alrededor que la usan como habilitadora.
Una pena, ¿no?
Mindset siglo XX

[Jose Manuel de la Chica @Delachica]

Artemis II left Earth on April 1. Mythos arrived on April 7.
A model autonomously found thousands of zero-day vulnerabilities. America's largest bank CEOs called an emergency meeting with the Fed.
Not because of a breach. Because of what a model can now do alone.
My latest on Substack: https://t.co/G3EnPL2p3e

[Jose Manuel de la Chica @Delachica]

Happy World Quantum Day.

Today we celebrate a relationship that is quietly becoming one of the most consequential in science: AI accelerating quantum discovery. Quantum, in return, promising to reshape what AI can even attempt.

Entanglement, it turns out, is not just a physical phenomenon.

#WorldQuantumDay #QuantumAI #Entanglement

[Jose Manuel de la Chica @Delachica]

Paradox of the moment: the term AGI has never been more popular — or less useful.
When it felt distant, it worked as a north star. Now that it's arguably here, it's become a distraction.
The people actually moving the needle aren't debating whether this is AGI. They're measuring concrete things and getting back to work.
Worth reading alongside this 👇
https://t.co/128tsFMGWs

[Jose Manuel de la Chica @Delachica]

@ifpenuelas También. Es un problema poliédrico, pero lo que puede hacerse es hacernos trampas al solitario con la manera de abordarlo. Y más con un gap creciente entre los sistemas de control y el arte de lo posible con la IA.

[Jose Manuel de la Chica @Delachica]

The human-in-the-loop isn't governance. It's a highway checkpoint designed for a different kind of traffic. When agents operate in swarms — collective decisions at non-human speed — no committee will be fast enough to be in the loop. Link to the post → https://t.co/6lZF35Yt3O

[Jose Manuel de la Chica @Delachica]

Asimov solved AI governance in 1942. His robots failed not because the laws were wrong — but because external constraints can't govern systems that generate behavior internally. We're making the same mistake with agentic AI. Convergence Layer is live: https://t.co/S7Z1Zw5rkb