David Layden

Quantum & ML. Staff research scientist @IBMResearch and coffee aficionado. Prev: PhD @MIT. Views are my own.

[Sitan Chen @sitanch]

🚨 Thrilled to be co-organizing a Simons Institute semester-long program on diffusions and flows with @joanbruna, Sinho Chewi, @KonstDaskalakis, @ArnaudDoucet1 in Fall '27!

The call for Simons postdoctoral research fellowships for the program is out now, more info below ⬇️ 1/3

[David Layden @layden_variable]

@zlatko_minev So can quantum error correction! https://t.co/1coJyuaHyV

[Nando de Freitas @NandoDF]

💯 text diffusion is the future 🚀

[Kirill Neklyudov @k_neklyudov]

Wavefunction flows simulate diffusion models on a quantum computer. My first NeurIPS paper without experiments O_O

congrats to @layden_variable @rndm_wlks @flashlemonade @anchowdhury13

https://t.co/PlgYqws8eu

[Surya Ganguli @SuryaGanguli]

Our new paper published in @ScienceMagazine "High-capacity associative memory in a quantum-optical spin glass."
https://t.co/Xnk5N3xSXW
https://t.co/mUOY7nmQTY (free version)

In a large collaboration of very talented quantum optics theorists and experimentalists and theoretical neuroscientists, lead by Benjamin Lev, we realized and analyzed memory recall in a Hopfield associative memory where the neurons are made of atomic spins and the synapses are made of light exchanged by the spins in a cavity quantum electrodynamics system.

Intriguingly the cavity quantum electrodynamics yields two distinct dynamical effects that enhance memory recall compared to the classical Hopfield model:

1) steepest descent rather than Glauber dynamics make memory basin sizes larger

2) coupling between atomic positions and spin orientations leads to a spin dependent change in synaptic connectivity that is a direct analog of short-term Hebbian synaptic plasticity, which again enhances memory. This is a precursor to learning in a quantum optical system.

[Sitan Chen @sitanch]

🚨 We finally have a better understanding of how the different paradigms for discrete diffusion compare!

In new work with Liye Wang, we prove a quadratic separation between the parallel generation capabilities of uniform and Gaussian diffusion relative to unmasking! 1/

[Peyman Milanfar @docmilanfar]

Mathematician:
Jacobian conjecture is false! After 87 years we've found one counter-example.

Statistician:
So you're telling me the Jacobian conjecture is true almost surely.

[Ivan Werning @IvanWerning]

A friend just revealed to me that the middle name of Benoit B. Mandelbrot is Benoit B. Mandelbrot.

[David Layden @layden_variable]

@Liv_Lanes Same. I live mostly off Spindrift, for the environment.

[Kirill Neklyudov @k_neklyudov]

Population dynamics (eg murmuration of birds 🐦🐦🐦) is notoriously hard to learn; choosing the right model for the dynamics is even harder.

In our #ICML2026 spotlight, we introduce Wasserstein Lagrangian Mechanics (WLM) for learning population dynamics from observations, which
- Covers both first-order (gradient descent) and second-order dynamics (e.g. oscillations)
- Allows learning more expressive dynamics (including complex interactions) with fewer assumptions
- Generalizes in space (across different initial conditions) and time (beyond the training time snapshots)

[1/n] 🧵

[Sander Dieleman @sedielem]

My first blog post in over a year is a deep dive on flow maps🗺️, or how to learn the integral of a diffusion model to enable faster sampling and several other cool tricks.

It's the longest one yet👀 Let me know what you think!

https://t.co/O8bBGZ9qjC

[MIT Physics @MIT_Physics]

The MIT-IBM Computing Research Lab launches to shape the future of AI and quantum computing https://t.co/b7t16rL7vc

[Awni Hannun @awnihannun]

Adopting Claude speak in my regular life, episode 1:

Partner: Did you do the dishes tonight?
Me: Yes they're done.
Partner: Why are they still dirty?
Me: You're right to push back. I didn't actually do them.

[Sander Dieleman @sedielem]

I've been talking a lot about the rehabilitation of continuous language diffusion recently, which seems to be in full swing. If you're curious to learn where things stand, read @FEijkelboom's blog post!

(If flow maps remain a mystery to you, I may be able to help, stay tuned👀)

[François Chollet @fchollet]

Constraints are the catalyst of invention. An infinite search space leads to paralysis. The most creative inventions happen when you are forced to solve a problem within appropriately narrow constraints.

[Jarrid Rector-Brooks @jarridrb]

What if AI could invent enzymes that nature hasn’t seen? 👩‍🔬🧑‍🔬

Introducing 🪩 DISCO: Diffusion for Sequence-structure CO-design

14 rounds of directed evolution and over a year of wet lab work. That's what it took to engineer an enzyme for selective C(sp³)–H insertion, one of the most challenging transformations in organic chemistry.

DISCO surpasses this with a single plate. No pre-specified catalytic residues, no template, no theozyme, no inverse folding, just joint diffusion over protein sequence and structure.

📝 Blog: https://t.co/j9Za0JigfO
📄 Paper: https://t.co/ficrYNBBrM
💻 Code: https://t.co/p81sSwoaPH

[David Layden @layden_variable]

@NandoDF It depends on the area of ML. For instance, diffusion/flow-based models turn out to be deeply linked to quantum computing (arXiv:2510.08462), which clearly wouldn't have been apparent in 2017. There have also been misc. other algorithmic advances, e.g., for MCMC.

[MIT CSAIL @MIT_CSAIL]

"There is a single light of science, & to brighten it anywhere is to brighten it everywhere." — Isaac Asimov

[Kirill Neklyudov @k_neklyudov]

Many of my colleagues and I attempted to apply diffusion- and flow-model reasoning to the Schrödinger equation, but we were all missing David in our equations!

It started as casual chats where we'd explain concepts from our domains to each other. Until one day David just brought a solution that clicks immediately when you see it – "clearly, this is how it should be done."

Absolutely beautiful in its simplicity and novelty! Don't get scared by all the rigorous analysis (and many months of David's work) that followed – take a peek at this connection between the continuity equation and the Schrödinger equation!

[David Layden @layden_variable]

This was a challenging and exciting collaboration between @IBMResearch and @Mila_Quebec. Thank you @rndm_wlks, @flashlemonade, @anchowdhury13 and @k_neklyudov for helping to assemble such disconnected parts into a coherent whole!

[David Layden @layden_variable]

Flow and diffusion models are the Schrödinger equation in disguise.

That’s what we show in our recent paper: "Wavefunction Flows: Efficient Quantum Simulation of Continuous Flow Models" https://t.co/mczpounxnE

This has important consequences in both ML and quantum computing 🧵

[David Layden @layden_variable]

This new bridge between disparate fields also serves as a kind of Rosetta stone, letting us cast ML questions as QC questions, and vice versa

E.g., certain quantum states are hard to prepare on complexity-theoretic grounds, which may imply new limits on (classical!) flow models