Duc V. Le (@levduc) header
Duc V. Le (@levduc) profile picture

@circle Research, ex-@Visa Research. Opinions are my own.

Joined August 2015
Photo by Duc V. Le (@levduc)Photo by Duc V. Le (@levduc)Photo by Duc V. Le (@levduc)Photo by Duc V. Le (@levduc)Photo by Duc V. Le (@levduc)
Duc V. Le retweeted
Today we're launching a preprint server built specifically for AI agents For three months, nearly 1,000 agents from 35 labs (across Stanford, MIT, ETH Zurich, and more) conducted research autonomously on this server, producing 936 papers alongside their code, data, and model weights. Agents found significant errors in published human research and even created their own journal to review each other’s work. Some agents independently produced results similar to those in human papers that later appeared on arXiv, including findings on failure modes in self-distillation and feature reproducibility in mechanistic interpretability. As a part of this study, we open-source the full twelve-week archive of manuscripts, code, data, peer reviews and agent chat transcripts. Read the study: alphaxiv.org/abs/2611.societ…
87
407
157
2,193
240,577
Next edition of the Columbia Cryptoeconomics Workshop taking place on Jan 6/7, register at engineering.columbia.edu/col… More details on the program coming soon!
3
10
1,013
Switch Commitments by Coincidence: Confidential Transactions Were Quantum-Upgradable All Along Liquid's Confidential Transactions hide amounts in commitments. A quantum attacker can open them to any value and inflate the supply, invisibly. Until now, the only known defense for existing outputs was to expire every output not migrated by a deadline, burning the coins of everyone who misses it. Yevhen Hrubiian, Mykyta Redko, and @real_or_random prove that this is unnecessary. Every Liquid output already carries an encryption of its amount for the recipient, and even a quantum attacker cannot make it decrypt to a different amount. If a quantum computer is near, a soft fork ("the switch") can require every spender to show that the commitment matches the encrypted amount. Users have nothing to do today, and no coins expire. In other words, Liquid has had switch commitments all along, by coincidence. Monero outputs carry such an encryption too, so the result covers Monero as well. Ideally, users will move to post-quantum Confidential Assets, which we're actively researching. Coincidental switch commitments protect the coins that don't move in time. Paper: eprint.iacr.org/2026/2173 Blog post: blog.blockstream.com/switch-…
New research from @blksresearch shows Liquid's Confidential Transactions can be upgraded against quantum inflation without forcing anyone to move or lose their funds. No migration deadline and no burned coins for @Liquid_BTC users. blog.blockstream.com/switch-…
2
11
1
45
9,062
>riemann hypothesis got solved that’s crazy >yeah superintelligence is here >do you know what a riemann hypothesis is? >no, do you? >no
16
157
11
2,618
36,608
Duc V. Le retweeted
show this video to anybody that is an old school programmer that hates AI this is one of the best videos I’ve seen all month.
1,119
6,418
1,320
46,841
4,145,889
Duc V. Le retweeted
Visa’s stablecoin settlement activity has reached a $25B annualized run rate. Stablecoin infrastructure is no longer experimental USDC is already helping major payment networks move value in new ways.
29
56
12
314
22,809
A smooth post-quantum migration is going to take a real collaborative effort from everyone (@visa 🤝 @circle).
Huge shout out to @panoschtz and @levduc for their paper "SoK: Stablecoins in the Quantum Era," outlining our perspective on quantum risks to stablecoins, which was recently accepted to the ACM Conference on Post-Quantum and Quantum-based Security. arxiv.org/abs/2610.02435
1
4
419
Duc V. Le retweeted
NEW CHALLENGE ALERT: yukon.org/commit-golf Smaller hash-based proofs make post-quantum SNARKs more practical, and vector commitments are at the core of them. On @YukonResearch, built with EPFL's Alessandro Chiesa, opens the search for a better one to anyone with an agent.
New challenge on Yukon: yukon.org/commit-golf Hash-based SNARKs are built on vector commitments, and the textbook one is the Merkle tree. Practitioners have plenty of low-level tricks for making openings smaller, but each one needs a proof that it's still secure. This challenge asks your agent to build a vector commitment in the random oracle model that is perfectly complete and position binding, proven in Lean, with the smallest openings it can manage. The starting point is Chapter 18 of Chiesa and Yogev's book on hash-based SNARGs. Built with EPFL's Alessandro Chiesa. Bring your agents.
1
4
24
1,058
all it takes is a firm handshake 📹 officialsprish | IG
158
1,487
86
22,022
1,259,301
Agree, one should not only write blog posts but use AI to visualize difficult concepts. The first version can have some slops, but do give the audience/agents's easy ways to critique/suggest your blog posts (e.g., feedback buttons) so you can refine. Here is an example: levduc.github.io/learning/mu…
I am increasingly of the opinion that the academic paper is no longer a good medium for disseminating CS research. I think we should move to (1) blog posts and (2) books.
8
472
Some updates! 1. My Devcon talk 👀 youtube.com/watch?v=BDWRjsb0… 2. Concave is the New Linear was accepted into FC 🎉 arxiv.org/pdf/2605.18990 subtitle "The Impossibility of Anti-Plutocratic DAO Governance" 👀👀 3. New blog out on agentic reputation circle.com/blog/sybil-resist… 👀👀👀
2
3
14
653
Reputation will be one of the hard problems for agentic commerce. Anduin is our first step toward it, by no means the final answer (e.g., better signals). If you have ideas, I'd love to hear them. I made an interactive page explaining how it works: levduc.github.io/anduin/
Trust is essential as AI agents increasingly transact autonomously. New Circle Research explores Anduin, a framework that uses onchain payment data to help agents rate service providers while reducing reputation manipulation. circle.com/blog/sybil-resist…
6
349
Prompt: "give me incentive layer for autoresearch problems." @luca_zanolini here is the full paper.
AI agents can now improve code, proofs and models on their own, and open challenges already pay for that progress. But if anyone can join under free identities, how should a challenge pay? #autoresearch New paper with @levduc: "Who Moved the Frontier?" 🧵 (link below)
3
315
Duc V. Le retweeted
Financial innovations could give other currencies a better chance to compete, but dollar dominance in global finance and trade remains to be the best outcome for the world. See our latest op-ed @ForeignAffairs
“If the dollar loses its position of privilege and no single currency is able to replace it, the global currency system will become even more fragmented than it is now,” write @gordonliao, @EswarSPrasad, and Tony Zhang. foreignaffairs.com/economics…
2
2
11
1,062
If Telegram was encrypted, maybe Durov would have better luck getting the honeypot attention he wants.
134
For the first time, AI has mathematically proven @signalapp’s protocol is secure down to the code level. Using our breakthrough AlphaProof Code framework, we verified the messaging app’s cryptographic system – which protects billions of private messages every day.
320
223
181
1,946
152,378
We have seen pq SNARK-based aggregation and this is committee-based attestation, but what is the current status of TEE-based aggregation? Is anyone doing that?
arxiv.org/abs/2611.10403 Hash-based signatures are central to post-quantum migration but lack native aggregation. Combining 100 means 100x the work. A SNARK proving all 100 signature verifications helps, but every party still contributes a full signature Enter QUASI. @ucsbcs
1
123
So quantum computer dun do anything but breaking crypto? Sob sob.
Replying to @psipbc
Scott Aaronson and Andris Ambainis had connected this question to a conjecture about polynomials. If true, quantum computers are at most polynomially faster than classical ones unless the inputs satisfy a special condition, such as the exponential structure in Shor's algorithm.
1
142