The N3XT Generation is Building the Future of Digital Ownership

Before You Dive In:

• Ledger N3XT asked university students to publish original research on digital ownership.
• The entries tested claims rather than simply surveying a field: authors ran experiments, built working hardware and audited a trading agent.
• Many asked how an AI agent that holds value should be authorized, audited and held to account, and several challenged Ledger’s own answer.

The Ledger N3XT research competition ran from August 4 to September 4, requiring students to submit original research on digital ownership and offer a candid analysis of how their work aligns with Ledger’s mission. Twenty-three of these student papers are now published in full on Ledger Academy. This recap explores the competition’s scope, key insights, and the three track winners who travelled to Paris for Ledger Op3n.

What Is Ledger N3XT?

Ledger N3XT is Ledger’s long-term program with nearly 70 university blockchain clubs worldwide, built to support the students who will shape digital ownership next with education, access to Ledger’s experts and a platform for their work. The research competition is the latest piece of that.

Why the Competition Exists & What It Asked For

Students are doing serious research in this space, but most of it rarely reaches the people building it. The competition was designed to change that: give students a visible home on Ledger Academy so their voice is heard in the industry, with their name, their university and Ledger’s reach behind it.

Between August 4 and September 4, 2026, students were invited to submit an original research paper on one of three tracks: the Agentic Economy, Privacy, and the Open track, covering any subject within Ledger’s mission.

Every paper had to be original works with a dedicated Ledger Lens, explained below. Authors then shared their papers publicly on X or LinkedIn, tagging @Ledger.

Papers that met these criteria were published on Ledger Academy. Among them, the winner in each track was the paper whose submission post had the most likes three days after the deadline. Each winner earned a seat at Students at Ledger Op3n in Paris on October 15-16, with travel and accommodation covered.

By the Numbers: Scale, Quality & Breadth

Most of these papers put a claim to the test rather than surveying a field.

• 23 papers live on Ledger Academy, across 3 tracks: Agentic Economy, Privacy, and Open tracks.
• Students from 7 countries (the US, Switzerland, Germany, France, India, Pakistan, and Nigeria)  across 4 continents
• 6 papers from the Oregon Blockchain Group alone

Authors come from the University of Oregon, USC, the University of Michigan, Purdue, San Diego State and CU Boulder in the US; EPFL, TUM and Kedge Business School in Europe; IIT Delhi and COMSATS University (Lahore campus) in Asia; and the Federal University of Technology, Owerri and the University of Lagos in Nigeria.

The subject matter goes just as far, from zero-knowledge proving systems and Ethereum finality to stablecoin payments, tokenized assets and onchain social benefits. A critical question surfaced across numerous papers: what mechanisms are required to properly authorize, audit, and hold value-bearing AI agents accountable?

Meet the Three Winners

Charles Sneed, University of Oregon (Agentic Economy track) — An Autonomous Agentic Trading Agent on Solana 

Sneed audited Omo, a Solana trading agent whose public portfolio climbed past $200,000. Before every trade, Omo wrote its reasoning onto the blockchain, a timestamped record no one, including the agent, could edit. Sneed checked that claim against the chain. The lock held, but the transaction paired with the decision was a token transfer Omo never signed, a signing-key gap highlighted in the study. His takeaway:

“A timestamp and a cause are separate claims.”

---

Drew Manley, University of Oregon (Open Track) — A Blueprint for Patient-Held Medical Credentials 

Manley took on a problem every patient knows: proving something about your health without handing over your records. His blueprint uses zero-knowledge proofs so a person shares only the fact a verifier needs, and argues the patient’s credential secret should be “held in dedicated hardware instead of inside a phone app.”

---

 Ivan von Greiff, TUM Blockchain Club (Privacy Track) — Inference of Attacks by Bandits Against Zero-Knowledge Virtual Machines

Von Greiff turned to the machinery behind zero-knowledge proofs. A zkVM can only vouch for a computation if its constraints capture every rule, so he built a fuzzing framework, tested on RISC Zero, that hunts for soundness bugs: flaws that let a broken computation pass as a valid proof. His Ledger Lens draws a line worth remembering:

“Hardware anchors authority, while sound verification anchors computational validity.”

---

All three winners joined the Students at Ledger Op3n cohort in Paris on October 15-16, with travel and accommodation covered, and met Ledger’s team and leadership in person.

Beyond the Winners: Three Tracks, Three Takes

The Agentic Economy Track

How do you let an AI agent act without letting it act alone? The authors addressed this concern from several angles.

Youssef Jeddi of the EPFL Blockchain Student Association removed the language model from a working system’s authorization path, replaced it with a deterministic policy engine and attacked both versions with 34 adversarial cases. His paper is built around a single question: when an agent transacts, “who said yes, and what does that yes prove.” 

Huzaifa Bin Hamid of COMSATS University Islamabad, Lahore Campus, argues in Signed, Not Sound that hardware-backed custody alone does not complete authorization, because a manipulated instruction can still become a validly signed payment.

The Privacy Track

Another paper that stood out came from Arnav Panjla and Sahitya Shankar of IIT Delhi, where they built an open-source hardware signer for about $10 and found its random number generator returns a fixed test pattern until permanently locked. Their point in The Custody – Privacy Gap is blunt: a key kept on a chip is custody, and “it is not privacy”. 

Austin Lee of USC ran 60 scenarios for Before the Signature paper and found a wallet can keep every predefined warning and still be unable to say what a transaction will do. His takeaway: 

“Warning coverage, predicted outcomes, and external disclosure should therefore be reported separately.”

Cameron Coleman from Oregon Blockchain Group redesigned the plumbing of America’s food assistance program. Tokenizing SNAP Benefits imagines eligibility tokens, encrypted receipt proofs and stablecoin-backed payments, so people can prove eligibility once, without revealing their income.

The Open Track

Kyrian Alex from Federal University of Technology, Owerri, compared Ethereum’s pre-confirmations, the Fast Confirmation Rule and full finality across six dimensions in Faster Than Finality. His Ledger Lens concludes that:

“Consensus secures agreement among validators, not human intent.”

Gaurav Dev of the TUM Blockchain Club brought a derivatives desk’s toolkit to Keep the Carry, Hedge the Tail, designing funding-rate options to insure the risk of perpetual futures.

Sebastian Sampedro (USC) argues that institutions are adopting the architecture of tokens while declining the principle of ownership. Thuy Duong Do (Kryptosphere) benchmarks stablecoins against SWIFT for cross-border SME payments. Loris Tran of EPFL proposes replacing recursive proof aggregation with accumulation in zkVMs. And two agentic papers refine their arguments into a single line each: Demilade Ayeku’s (University of Lagos) Keys Have Lives reminds us that “A key is not an event,” while Samuel Smith’s (University of Michigan) The Shrinking Anchor observes that “An anchor holds whatever it is attached to.”

The Ledger Lens: Analytical Perspective on Our Mission

Every paper had to include a Ledger Lens: a section where the author looks at their findings through Ledger’s mission and says, honestly, where self-custody and secure hardware help and where they fall short. It was never meant as a pitch. Security only improves when someone is allowed to say where it breaks, so we asked them to test our assumptions and published the results. Several did exactly that.

Sneed’s cautions that secure hardware would fix one problem, binding a signed trade to a locked decision, but not the rest. Diego Carpenter of the University of Oregon argues in The Stag Hunt Problem that Ledger’s self-custodial model is necessary but not sufficient for agent authorization, and The Shrinking Anchor warns that hardware verification could cover a smaller share of agentic transactions as volumes grow. That adversarial instinct is the same one that drives the Ledger Donjon.

Read all the papers in the Ledger N3XT series on Ledger Academy.

What Comes Next?

AI agents now read inboxes, browse the web and increasingly act without anyone confirming each step. The people who will build the safeguards for that world are sitting in lecture halls and club meetings today, and Ledger N3XT exists to meet them there, with education, access to experts and a platform on Ledger Academy. Their questions, about who authorizes an agent and what a signature actually proves, were the ones at the center of Ledger Op3n 2026.

When an AI acts for you, the last word should still be yours. Judging by these papers, the next generation is already working out how.

Read all the papers in the Ledger N3XT series on Ledger Academy. Want to bring Ledger N3XT to your university club? Get in touch at [email protected].

---

Disclaimer

Student research is published via the Ledger N3XT Research Competition. Findings are the authors’ own and are not necessarily endorsed by Ledger.