Andrés Miguel Torrubia Sáez

Biodrug design @ AI ∩ biology

[Mark Lewis, MD, FASCO @marklewismd]

Cheers, chills, and a standing ovation when RASolute 302 showed unprecedented survival on daraxonrasib for patients with progressive pancreatic cancer

Seldom do you sense you’re witnessing a historic moment in cancer care but this feels like ras targeting has arrived

#ASCO26

[Son Luong @sluongng]

Codex just found a “workaround” of not having sudo on my pc…

[Andrés Miguel Torrubia Sáez @antor]

@cortesi you mean 5.5 right? RIGHT? 😂

[David Bonilla @david_bonilla]

En 2018, @antor dio una charlaza sobre IA en la @tarugoconf en la que avisaba de lo que está pasando hoy, 8 años más tarde, y cómo podíamos aprovecharlo.

Espero que las charlas de la #TRG26 nos ayuden a vislumbrar cómo será el sector en 2034.
https://t.co/j4cdVTeMoR

[Andrés Miguel Torrubia Sáez @antor]

Me hizo muchísima ilusión dar esa charla y si me hubieran dicho entonces que en el 2026 estaría usando la IA para desarrollar fármacos no me lo creería pero...

TODO ES POSIBLE

[Eduardo Manchón @eduardomanchon]

My hot take:

1/ Most AI-email startups will fail. Every YC batch ships 3 new ones. Brilliant tech, beautiful UX, but building "a better inbox" and asking people to migrate has been a losing bet for 20 years. AI doesn't change that.

2/ The misunderstanding is thinking AI is about productivity: replying faster, summaries, drafts, daily briefings, integrated calendars... it's all just meh. The AI revolution is much more than the inbox zero mentality.

3/ Think about your inbox as a server where every email is a log entry: sender, timestamp, subject, plus a wall of text starting with "Hi Joe, hope you're doing well..." Suddenly the entire format feels primitive. Obsolete. We're still using a protocol designed 40 years ago.

4/ The fix isn't a new inbox. That's why at Mailsuite we're not trying to replace Gmail or rebuild the inbox with AI. We're building an intelligence layer on top of the inbox people already use. No migration. No learning curve. Our game is different.

5/ Mailsuite AI processes your inbox data, identifies what's relevant to you and writes it back in human language.

6/ Why Mailsuite: 14M signups. 800k monthly active users. 119k paying customers. 4 billion emails tracked. All organic, no paid marketing, no sales team. We have the users, we have the data, now we have the AI.

7/ Email hasn't really changed for 40 years.What we're about to ship will. Soon.

[Andrés Miguel Torrubia Sáez @antor]

@JaimeObregon Para dictar, con modelo local Parakeet V3
https://t.co/F0UJT4g41I

[Andrés Miguel Torrubia Sáez @antor]

I beg to disagree re: loudspeakers.

~1970 => The best sound good.
>2000 => They best sound real.

The reason why avg person does not know about them is more about market dynamics and preferences rather than technology peaking.

[Andrés Miguel Torrubia Sáez @antor]

@XMihura hace ~2 años, un chaval de 18 años en una de las unis top USA haciendo matemáticas me dijo que había llegado a la conclusión de que había nacido un siglo demasiado tarde para poder hacer algo y que full pivot a IA. hoy: dropout y lo acaban de fichar en oai

👀¿la tesis pa qué?

[Andrés Miguel Torrubia Sáez @antor]

@cortesi The only reason I got 'Ultra', to use Deep Think results in "Too busy right now, try again later" >50% of times.

Way too unreliable.

[Aldo Cortesi @cortesi]

So the situation didn't improve with 3.5 Flash and the new Antigravity - in fact, it got worse. Either Google can't compete or it won't compete in coding. Both answers are fascinating in completely different ways.

[Andrés Miguel Torrubia Sáez @antor]

A bad day to be an AI negationist

[OpenAI @OpenAI]

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946.

For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids.

An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better.

This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.

[Shunzhi Wang @wang_shunzhi]

A growing protein cage has no architect.

Each piece only responds to its local environment.

Yet somehow, thousands of pieces can come together to form a large, precise sphere.

Today in @Nature, two back-to-back papers from our team and @UWproteindesign show how de novo protein design can make this kind of self-assembly programmable.

The trick: we didn't design perfect symmetry.

We designed the rules for breaking it. [1/17]

https://t.co/zHeNBc1oFt
https://t.co/MNS1bS9r5O

[Arturo LoAIza-Bonilla, MD MSEd @DrArturoAI]

We need benchmarks like frontier models to prove the gains @demishassabis - the real humanity’s last exam is to get them optimized, Pareto frontier included @arcprize @biohub

[Andrés Miguel Torrubia Sáez @antor]

@niizan_AVC I ❤️ both FW900 + W900

[Andrés Miguel Torrubia Sáez @antor]

@inakitajes ¿ejemplos concretos?

[Andrés Miguel Torrubia Sáez @antor]

I wish the 🇪🇺 had leaders like Singapore's Minister for Foreign Affairs 👇🏻

[Vivian Balakrishnan @VivianBala]

Built a personal AI agent - why I did that and what it means. https://t.co/lo8SLUMD9s

[Andrés Miguel Torrubia Sáez @antor]

Biology is so vast that:

a) collectively we still can't reliably tweak it to cure disease X without causing side-effect Y

b) individually no biologist holds more than a sliver

That's the opening for AI to reinvent biotech.