5 Published Papers

Research that becomes real infrastructure

We publish first, build second. Every Cryptuon product traces its lineage to a peer-reviewed paper with formal guarantees.

Publishing before building is not an academic affectation. A protocol paper forces the safety and liveness properties to be stated before any code depends on them, which is the only point at which those properties are cheap to change. By the time a relayer, an SDK, and three integrations exist, a missing guarantee is a migration rather than an edit.

The five papers below are not equally weighted. Two of them — universal atomic composability and fair transaction ordering — define constructions that recur across the portfolio, and the commit-reveal primitive from the second appears in cross-rollup swap ordering, in sealed-bid auctions, and in the collusion-resistant scoring used by the deepfake detection network. The other three are framing work: they describe a class of system and the incentive traps it falls into, and they shape which products exist rather than how a specific one is implemented.

Each entry names the product that carries the result into production, what the paper actually changed about the implementation, and a technical write-up that goes deeper than an abstract can. Where a paper has not yet produced shipped code, we say so instead of implying otherwise.

a16z Crypto Startup School Powers: Tesseract

Towards Universal Atomic Composability

A formal framework for achieving atomic transaction execution across heterogeneous rollup environments. Introduces a three-phase coordination protocol (Buffer-Resolve-Execute) with provable safety and liveness guarantees.

What it changed in the code

The paper treats a cross-rollup swap as a distributed-commit problem rather than a messaging problem, which is why the protocol reasons about buffering and resolution rather than about relayers and proofs of delivery. Buffer-Resolve-Execute keeps participants in a state where either every leg commits or none does, so a partial fill cannot strand assets on one rollup. Tesseract is that protocol implemented in Vyper with a Rust relayer, and the commit-reveal ordering step in the same design also supplies the MEV protection described in the FairFlow work below.

Further reading: Why Multi-Chain DeFi Breaks: Cross-Chain Atomicity

Cross-chain Formal Methods Rollups

Generalised DePIN Protocol

A generalized framework for decentralized physical infrastructure networks that aligns economic incentives between hardware operators, protocol maintainers, and end users through mechanism design.

What it changed in the code

Every physical-infrastructure network faces the same bootstrapping trap: operators will not buy hardware before there is demand, and demand will not arrive before there is capacity. The framework separates the subsidy that solves that trap from the steady-state fee that has to survive once the subsidy ends, which is the distinction most DePIN token designs collapse into a single emission curve. Four projects in the portfolio inherit this structure directly, including the GPU-backed inference network and the indexing and storage networks on Solana.

Further reading: Decentralized LLM Inference with Solana Settlement

DePIN Infrastructure Mechanism Design

Powers: Tesseract (MEV protection)

FairFlow Protocol

An equitable MEV mitigation protocol ensuring fair transaction ordering across block producers. Introduces a commit-reveal mechanism that eliminates front-running without sacrificing throughput.

What it changed in the code

Fair ordering is usually framed as a constraint that costs throughput. The paper argues the opposite: a commit-reveal window adds latency but removes the need for the defensive gas auctions that consume block space in the first place. The primitive it formalises is small enough to be reused, which is why the same commit-reveal construction now appears in three unrelated places in the portfolio — cross-rollup swap ordering, sealed bidding, and the collusion-resistant scoring used by the detection network.

Further reading: Verifiable On-Chain AI Without a Trusted Oracle

MEV Fair Ordering DeFi

Decentralized Deepfake Detection Network

A coordination protocol for distributed media verification. Connects verification requesters with independent detection node operators through economic incentives and reputation-weighted consensus.

What it changed in the code

The paper argues that the weak point in media verification is not classifier accuracy but correlated failure: when a thousand organisations query one vendor they hold one opinion, replicated, and they all inherit the same blind spot on the same day. A network of independently chosen models turns that into an ensemble, and recording the dissent alongside the verdict lets a downstream system treat a 5-0 consensus differently from a 3-2 split. DFPN is the Solana implementation, with stake, graduated slashing, and an on-chain audit trail.

Further reading: Deepfake Detection as a Decentralized AI Network

AI Safety Content Verification Reputation Systems

Centralized Intermediation in a Decentralized Web3 Economy

An economic analysis of how centralized intermediaries emerge in ostensibly decentralized ecosystems, with proposed mechanisms for preserving decentralization properties at scale.

What it changed in the code

Intermediaries reappear in decentralized systems wherever a task has strong returns to scale — indexing, relaying, sequencing, custody. The analysis treats that as an equilibrium rather than a failure of will, and asks which mechanisms keep the resulting intermediary contestable instead of trying to legislate it away. That framing is why several portfolio projects pay independent operators to compete on a measurable service level rather than assuming volunteers will materialise, and why the tokenomics differ so much from project to project.

Further reading: Tokenomics Across Cryptuon: Earned vs Allocated

Economics Decentralization Web3

Our Approach

From paper to production

A rigorous pipeline that turns theoretical breakthroughs into shipped software.

Identify Problem

Find a real infrastructure gap that existing solutions don't address.

Publish Research

Formalize the solution with provable guarantees and submit for peer review.

Build & Test

Implement in Rust or Zig with comprehensive test suites and documentation.

Ship Open Source

Release under MIT license. Build community. Layer enterprise features on top.

Open problems we are still working on

Liveness under relayer failure. The atomic composability protocol proves safety without trusting the relayer: a crashed or hostile relayer cannot cause a partial settlement. Liveness is weaker. A swap can stall until a timeout expires, and while no participant loses funds, capital is locked for the duration. Shortening that window without weakening the safety argument is unsolved.

Pricing a commit-reveal window. Commit-reveal removes front-running by hiding intent until ordering is fixed, at the cost of a mandatory delay. How long that window should be is currently a governance parameter set by judgement. We would like it derived from measurable properties of the venue rather than chosen.

Reputation that survives identity churn. Staked reputation works while stake is expensive relative to the gain from misbehaving. It degrades when an operator can abandon an identity cheaply and rejoin clean. Making the cost of a fresh identity scale with the value at risk, without excluding honest newcomers, is an open mechanism-design question across every operator network in the portfolio.

Benchmarks that cannot be gamed. Any network that pays operators by measured quality creates an incentive to optimise for the measurement. Hidden, rotating evaluation sets under governance are the current mitigation. A construction where the benchmark is verifiable without being predictable would be materially better.

If you work on any of these, or want to build on the published results, the contact page reaches us directly. Implementations of every result live in the product portfolio, and the technical write-ups are collected under Insights.

Want to collaborate on research?

We're always looking for research partners, academic collaborators, and teams building on our published work.

Reach Out Sponsor research