Iftach Yakar

@quantumyakar.bsky.social

Quantum computing + information | Computer science | quantum software engineer @ Qedma quantum computing

New paper from our team: QESEM on IBM’s Heron processor enabled precise measurements in a regime where leading classical simulations become unreliable. Main results were corroborated on Quantinuum systems. This is a step toward quantum processors as scientific instruments. arxiv.org/abs/2607.24937

Resolving Structure in Prethermal Floquet Dynamics with Precision Quantum Computation

Periodically driven interacting quantum many-body systems can exhibit long-lived prethermal dynamics, where local observables retain coherent structure even as entanglement and operator complexity gro...

arxiv.org

Reading “Project Hail Mary” after watching the movie and I wonder whether we could use astrophage to help build a quantum computer. They did mention it gives rise to Majorana particles, so maybe that’s one way?

I gotta admit, I was cynical about attending our MSc ceremony. After all - I have the degree already, so what use is there for dressing up and receiving a certificate on stage? But seeing how my kids were proud of their dad, and how it inspired them was a great feeling indeed.

BildBildBildBild

4yo: where did all the dinosaurs go? Me: an astroid hit earth and they all died. 4yo: ? Me: … *realizing she meant the dinosaur toys that were on the windowsill which my wife had moved* oh

Been messing around with creating MCPs that let LLMs run quantum tools like Stim or Microsoft QRE. I wanna see if we can get reliable and interesting results, even for exploration. If you favorite LLM would know how to use these, what would you ask it? Curious to get any feedback.

Window shows Claude chat the uses the Stim-MCP.
######
Left side (chat):

show me how the logical error changes as I use larger and larger surface codes.

Mar 17
I'll build rotated surface codes for distances 3, 5, 7, and 9, inject depolarizing noise, and sample logical error rates. Let me run them one by one.


Create circuit
Create circuit
Let me use Stim's proper surface code format with detectors and observables so we get actual logical error rates:


Used stim-quantum integration
Used stim-quantum integration
Good, d=3 logical error rate ≈ 0.622%. Now let me use Stim's built-in surface code generator for cleaner, properly structured circuits for all distances:


Used stim-quantum integration
Used stim-quantum integration
The X-ancilla measurement rates ~50% means the H gates are putting them into superposition before measurement without proper reset into the right basis. Let me take the cleaner approach — use the well-known Stim surface code circuit format that's guaranteed correct:
...

######
Right side (preview):
*shows some output plots from later in that conversation*
Surface Code Scaling — Simulation vs. Google Hardware, with Willow-matched noise
Stim v1.15 · rotated_memory_z · MWPM decoding (pymatching) · 10 rounds · 200k–2M shots
Google 2023: Sycamore · Nature 614, 676–681  |  Google 2024: Willow · Nature Dec 2024 / arXiv:2408.13687

In quantum information theory, we don't say "I love you", we say "Our correlation is so strong that the Coffman-Kundu-Wootters inequality prevents me from being entangled with any other subsystem", and I think that's beautiful.

Bild