Anthony Leverrier

researcher on quantum error correction

[Anthony Leverrier @letonyo]

Certainly the most surprising result in my career so far: a new qLDPC decoder that I thought had no chance to work!

We call it the Frontier decoder, joint work with Rüdiger Urbanke (1/4)

[Anthony Leverrier @letonyo]

@_cweedbrook @WhiteHouse @XanaduAI congrats!

[Anthony Leverrier @letonyo]

@olivez good catch! I had not seen that paper.

[Anthony Leverrier @letonyo]

Wait, Scott Aaronson is an investor?

[Oratomic, Inc. @TeamOratomic]

https://t.co/82Pi8pEel9

[nick farina @nick_farina]

what's notable isn't @teamoratomic raising $300M given their caliber of talent, but WHO they raised it from. a lot of firms on this list who long said QC was too far off to be worth major investment.

that changed today. QC is now in the domain of the major funds.

https://t.co/A23xXjU17f

[Anthony Leverrier @letonyo]

@nick_farina thanks for the explanation! I wasn't aware of this.

[Anthony Leverrier @letonyo]

@kimmonismus why would you still use Opus if you have access to Fable?

[Anthony Leverrier @letonyo]

is it worth using a codex reset 12 hours before the normal reset? I guess it just costs me a full reset for 12 extra hours, or am i missing something?

[Anthony Leverrier @letonyo]

https://t.co/Z40T5h7L6a

[Anthony Leverrier @letonyo]

If you're using the Frontier decoder to optimize and benchmark your qldpc circuits, you're the 1%.
We're still early.

If you're not, just ask your agents to give it a try.

[Anthony Leverrier @letonyo]

@kimmonismus next week seems so far way when claude's users have access to fable

[Anthony Leverrier @letonyo]

@MathdeProf @MoutonMatheux pour la question 1 du brevet, ça promet pour le reste

[Anthony Leverrier @letonyo]

am i the only one who sees this:
- me: "codex, how long would it take to implement this idea?"
- codex: "1-2 days for the simpler version, likely a couple of week for a more complete implementation"
- me: "ok do it"
- codex, 20 minutes later: "done!"

[Anthony Leverrier @letonyo]

@WoodsMischa oh i think this is indeed pretty common unfortunately.

[Anthony Leverrier @letonyo]

I've also been gradually switching to this value-driven way of working these past few years, without articulating it as clearly as Arvind.
But it certainly helps to have a tenured position to try this.

[Arvind Narayanan @random_walker]

At the start of my research career I operated in a deadline-driven mode because that's what most researchers seemed to do. Gradually I discovered the value-driven way of working. I'm glad I had a supportive advisor who didn't make me chase deadlines. It took me 20 years to fully embrace the switch — it requires developing a long-term vision, willpower to create structure without deadline pressure, a theory of value, project management skills, good taste, the willingness to turn projects down, brutal honesty about whether our work is any good (even if it gets published), and a lot more. But there is no going back!

[Jelani Nelson @minilek]

Update: I've joined @AnthropicAI and taken leave from the university. Excited to work with many talented, mission-driven people on the defining technology of our time.

[Anthony Leverrier @letonyo]

Really eager to try the Frontier decoder in this context !

https://t.co/Z40T5h7L6a

[Michael J. Biercuk @MJBiercuk]

@sarah_schupp No theoretical decoder survives first contact with laboratory errors

🤣

[Victor Reis @vetohaze]

In 1990, Michel Talagrand showed that given a₁,…,aₘ ∈ ℝⁿ, there are weights w₁,…,wₘ ≥ 0, only O(n log n) of them nonzero, such that ∑ᵢ wᵢ|⟨aᵢ,x⟩| ≈ ∑ᵢ |⟨aᵢ,x⟩| within 1% for all x ∈ ℝⁿ.

36 years later, that last log is gone: O(n) nonzeros suffice! [1/5]

[Anthony Leverrier @letonyo]

If you have specific codes, circuits or noise models you'd like to investigate, you should try the Frontier decoder. We only report results for surface, color and BB codes in the paper. But I've tried it for many other settings, and it works great.
https://t.co/JTvdmj2QjD

[Isaac Kim @Isaac__kim]

@letonyo For quickly comparing LER under different syndrome extraction circuits, across different codes. I found the alternatives to be slow/hard to tweak in comparison.

[Anthony Leverrier @letonyo]

Love it that Isaac is using the frontier decoder *frequently* even though it's only been available for less than 2 weeks ;-)

[Isaac Kim @Isaac__kim]

@letonyo I use it frequently and it’s awesome! It is often my go-to option.

[Anthony Leverrier @letonyo]

@Isaac__kim great! Yes, frontier is well suited for this problem. Thanks for using it!

[Anthony Leverrier @letonyo]

Most decoders for quantum LDPC codes hate degeneracy. It is the opposite for the frontier decoder: degeneracy means several paths can be merged, which reduces the frontier size.

You can test it on your preferred codes (github link below)