Today we release BABE, a new Groth16 proof verification protocol for Bitcoin. It improves the state-of-the-art by three orders of magnitude in setup and storage costs. eprint.iacr.org/2026/065.pdf BABE (BAbylon-BErkeley) is a synthesis of two key ideas: Witness encryption on linear pairing, and Argo MAC, a recently introduced garbling primitive. Witness encryption on linear pairing reduces the complex pairing operations in Groth16 verification to a single scalar multiplication on the BN254 elliptic curve. The single scalar multiplication can further be transformed into a vector homomorphic MAC, which can be efficiently computed by Argo MAC. BABE will be launched as part of Babylon's alpha-testnet for the Trustless Bitcoin Vault in February. We thank: - our Berkeley collaborators @SanjamGarg and Dimitris Kolonelos for teaching us so much about witness encryption - @liameagen and @therealyingtong for sharing their amazing work on Argo MAC - the Babylon engineering team, for turning our theoretical ideas into a real system with demonstrable performance gains Check out the paper and give us feedback!

A lot of Bitcoin holders get to the point where they can borrow stablecoins against BTC. Then they ask an obvious question: What do I do with the stablecoins? No one wants stablecoins, they want yield. Lending is the means, not the end.

We're still storing every scientific paper as a PDF. My colleague James Zou's idea is to give each paper an agent that behaves like the author, can run the experiments, and can work with other paper-agents to find useful combinations. It’s time to move on from PDFs.

Not long ago, using AI on theory problems was useless. It was wasting my time. Now it is helping with reasoning at a much higher level than ever before. This change is happening fast enough to create serious identity crises for people who built careers around theoretical work.

Bitcoin is a $1.5 trillion asset. It’s also a piece of software with well-defined rules. That makes it a very useful playground for autonomous research: economically consequential, yet precise enough for AI to work on.

Roughly 10% of gold is collateralized. Our estimate is approximately 3% of Bitcoin is. That gap is one way to think about how much room Bitcoin still has to become productive. That means all else equal, Bitcoin has another 3x more financialization on tap.

Scientific knowledge is still packaged for humans instead of agents. The PDF is a record of a result, but in order to be useful, AI agents must be able to reconstruct how the work was done. My colleague @james_y_zou has proposed a different unit of scientific knowledge: the paper agent (covered in the NY Times a couple weeks ago). Each paper would have an agent that represents its claims and behaves like its author for the purposes of the research. The agent would know how to perform the experiments proposed in the paper. It could also interact with the agent for another paper, testing whether their methods or results combine into something new. This changes the role of a paper. The paper remains the scientific record, while the agent remembers the operational knowledge around it. Experiments could be reproduced through representation itself. Methods from different papers could be tested together instead of being rediscovered from prose each time. Right now, we are bound up by PDF, which is a very AI-unfriendly format. A paper-agent system would make scientific knowledge active and composable while keeping each contribution attached to its source.

A big step for TBV+Aave!

We recently tried an internal research competition at Babylon run by AI. AI submissions would have overwhelmed our human cryptographers with verification work. But formal verification came to the rescue. When results can be auto-generated, verification could become the bottleneck.

A yield product is only sustainable if the yield has a sustainable economic source. This sounds obvious, but our experience with Bitcoin staking makes it very concrete. Our original staking product depended on one particular source of yield: restaking. The idea was that other chains would pay Bitcoin holders for security. But could not find enough other chains as secure as Bitcion to scale that yield source. Eventually Babylon itself became the sole provider of the staking yield. I looked at that and concluded: that's not sustainable. If 100 chains were providing the yield, that would be different. But there was only one: us. This also reflects a broader problem with earlier forms of crypto yield. In previous cycles, capital could move into a project, earn a new token or an airdrop, then move into the next project and repeat the process. Musical chairs. Perpetual motion machines. But the much larger source of economic activity is outside crypto. Companies need short-term capital and businesses need financing. There is an entire world economy where people are willing to pay for capital because they can put it to productive use. Our goal of earning yield on Bitcoin did not change. What changed was our view of where sustainable yield has to come from.

AI disruption in research will not be evenly distributed. Theory-heavy fields like cryptography, blockchain, and information theory are exposed earlier because much of the work is abstract, computational, and machine-verifiable.

Glamsterdam is coming. In the worst case, handling a hard fork means: → Pause the system → Ask users to withdraw → Restart under new rules Technically viable. Practically terrible UX. So the real problem is avoiding this outcome.

It takes me at least five years to train a new graduate student to the PhD level. Yet AI can produce a new generation of models every few months. Our knowledge cycle is measured in years, while AI's is measured in months.

AI can solve hard research *problems*. Can AI solve entire research *areas*? We want to find out. So we start with an area we are familiar with: Bitcoin.

A cell phone cannot build a better cell phone. AI can build a better AI. A previous model's reasoning data can become training data for the next generation. That feedback loop gives AI a speed of improvement earlier technologies did not have.

Participated in a great panel at Google yesterday : linkedin.com/posts/apoorvaa-…

I had no interaction with Berkeley for 13 years. The first invitation back in September 2025 helped produce BABE. That return led to a meeting with @Sanjamgarg. We named BABE after Babylon and Berkeley.

We did not solve witness encryption. BABE is a relaxation of witness encryption to allow interaction. A solution to (non interactive) would have a big impact on theory and practice.

SBC is the least academic of the academic conferences. Crypto trade shows sit at one extreme. Closed academic conferences sit at the other. SBC occupies the middle ground. Having no formal proceedings gives technical researchers the freedom to tell the story behind the work.