Working at the convergence of AI and Quantum Computing. Manager, Quantum+AI @IBMResearch. My views are my own.
Joined March 2011
- Tweets4.7K
- Following1.2K
- Followers957
- Likes5.6K
Juan Cruz-Benito retweeted
Last night an AI agent performed brain surgery on a mouse.
We gave it 19 gene therapy vectors in a spreadsheet, a mouse model of epilepsy, access to a motorize surgery rig with cameras and one goal: reduce seizures.
It picked the drug. It picked the brain region. It derived the coordinates from three papers. Then it drove the stereotaxic arms into the brain and injected 250 nanolitres while watching the entry site through the rig camera for bleeding.
A licensed vet gated every irreversible step and held a physical kill switch to stop the AI agent.
Thread on what it decided, what it saw, and the moment it refused to let me override it.
[Writing this on the spot out of excitement, but I've been 18hrs in front of my laptop coding and coordinating this, so keeping it brief and descriptive, more cohesive thoughts on this soon...]
Juan Cruz-Benito retweeted
BREAKING NEWS
The Royal Swedish Academy of Sciences has decided to award the 2026 #NobelPrize in Physics to Francis Halzen “for decisive contributions to the IceCube Neutrino Observatory and the discovery of high-energy neutrinos of astrophysical origin.”
Juan Cruz-Benito retweeted
Aprovechando el tirón de nuevos seguidores por @hackspain, aprovecho para compartir el Manifiesto de @exp8fellowship.
Para que se entienda mejor por qué existe Exponential.
Hello IEEE Quantum Week 2026!
Tomorrow, I’ll be giving a talk about our work in AI for Quantum Error Correction at the “AI for Circuit Synthesis, Optimization, and Discovery” workshop.
If you’re interested in our work at @IBMResearch or just want to say hi, don’t be shy!
I'm reading today some news about an LLM trained using Quantum Data and claiming being the 1st one ever.
Well... we published a paper 1 year ago post-training LLMs using Quantum Verifiable Rewards (RL using feedback from execution in QPUs or simulators) arxiv.org/abs/2508.20907
Juan Cruz-Benito retweeted
A bit ago I posted a follow up blog on my work automatically validating Qiskit code using Mellea. In the second blog I dug into some benchmarking and how functional validation could improve things further mellea.ai/blogs/qiskit-ivr-f…
New preprint with @ismaelfaro: "Qiskit QuantumKatas — a benchmark for evaluating LLMs on quantum code".
We translated Microsoft's QuantumKatas to Qiskit (350 tasks, 26 categories) and ran 16 models -frontiers, open weights- × 7 prompting setups = 39,200 runs.
Models implement known algorithms well (Simon's 82%) but struggle to encode problems into quantum terms (SAT+Grover's 34%). Most failures are bad quantum reasoning, not syntax. And chain-of-thought helps reasoning-tuned models but hurts the rest.
Paper arxiv.org/abs/2605.27210
Juan Cruz-Benito retweeted
I've been working on an open source project called Mellea, and wrote a blog post about using it to automatically validate and fix Qiskit code generated by an LLM: mellea.ai/blogs/qiskit-ivr-c…
Juan Cruz-Benito retweeted
Let me demonstrate the true power of llama.cpp:
- Running on Mac Studio M2 Ultra (3 years old)
- Gemma 4 26B A4B Q8_0 (full quality)
- Built-in WebUI (ships with llama.cpp)
- MCP support out of the box (web-search, HF, github, etc.)
- Prompt speculative decoding
The result: 300t/s
(realtime video)
Juan Cruz-Benito retweeted
Este post es buenísimo, además de enseñarte a navegar en una empresa grande y mantener una buena reputación también muestra uno de los mayores problemas de éstas. La política y los incentivos están por encima de lo que es bueno para la empresa
lalitm.com/post/why-senior-e…
✨We got AI agents to train RL models for quantum transpilation!
What if you could tell an AI assistant: "train a model to synthesize linear function circuits for X qubits on a ibm_nighthawk" - and it just does it?
The MCP server supports:
- Permutation synthesis (SWAP routing)
- Linear function synthesis (CNOT circuits)
- Clifford synthesis (H, S, CNOT gates)
This is how we envision training AI-powered transpilation models starting now.
#QuantumComputing #AI #RL #Qiskit #MCP
More info juancb.es/post/2026-qiskit-g…
Qiskit MCP servers: github.com/Qiskit/mcp-server…
The example agent: github.com/Qiskit/mcp-server…
