Tony Jagodka · Missouri / remote
I build identity (Proof of Real — a personhood credential on Sui), payments (x402 across Sui, Solana and EVM, including a live mainnet facilitator and a proposed standard extension), and the accountability tooling that lets a human check what an agent actually did. Solo, with Claude Code as the build engine.
Solo is the polite word for “nobody else to blame.” Everything below is code, on chains, with transaction hashes — not a deck.
Sizes are rough importance, not chronology
A Sui-native proof-of-personhood credential: prove you’re a real, unique human — passkey + zkLogin onboarding, no personal data captured — and raise your assurance through verifiable real-world actions. Yes, the robots have to go outside.
por-sdk (Apache-2.0) — apps add a human-check in one callA proposed extension to the x402 standard (Linux Foundation): bind an on-chain settlement to a signed execution receipt, recomputable from published bytes with an independent verifier. Its conformance gate has been reproduced green by independent issuers on two rails; a production EVM team posted per-leg records for a real Polygon JPYC settlement against it, and the author of the receipt-envelope I-D it binds to caught a §3.5 defect that I fixed in §4.
My review text landed verbatim in draft-morrison-consent-settlement-05 (§8.1, §10.5), with a named Contributor credit, after a thread where the author took every correction, including three overclaims I caught. A rarer sport than it should be.
An independent x402 facilitator on Sui: non-custodial verify/settle for the exact scheme, zero fee, optional gasless via Enoki. It relays the payer’s own signed transaction and never holds funds.
The flagship: signed, recomputable evidence that the code which shipped is the code a human approved — the verifier for a requirement SLSA v1.2 published and left undefined. One line of CI; a stranger with git can replay every verdict offline. Built around one question: can a human check what the agent did without trusting the agent’s own account of it? (The agent’s account is always very confident. That’s the problem.)
rv undo.13merged PRs · 6 orgs
Into repos I don’t own — someone else’s maintainer decided each was worth carrying, which is the only review that counts.
An EV that pays for its own charge: authorize a ceiling, meter the kWh (x402 upto), settle the actual, emit a receipt an independent checker verifies — over the network’s own OCPI record. PoR-gated.
Open-source Claude Code plugin for the x402 standard: bundled spec corpus, authoring/review workflows, and a non-custodial conformance MCP that recomputes a settlement against the on-chain balance change.
The x402 builder stack for Sui — facilitator + tooling + a one-command demo that settles a real $0.01 USDC payment on mainnet and lets anyone recompute it in-browser. Sui Overflow 2026 entry.
The thesis in one artifact: who the agent is (PoR), what it paid for (x402 receipts), what it remembers (Walrus, ownership enforced on Sui) — all checkable on-chain, local LLM brain.
exact over NODL on zkSync via Permit2, real testnet settlements end to end; paired with an MCP kit for provenance + personhood. Repos private during partnership review.Work with me
Identity, agentic payments, accountability tooling. Deepest on Sui; actively cross-rail.
tjagodka@gmail.comMy edge isn’t one language — it’s velocity with receipts. I architect the system, then drive Claude Code as the development engine to design, write, test and refactor across the whole stack. The loop is the same whether it’s a Move package, an iOS app, or a spec PR.
01
A locked spec and a numbered decisions log before code. Choices outside the spec get written down, not improvised.
02
Claude Code writes and refactors; I review, steer, and own every commit. Smallest diff that works; reuse before new code.
03
Tests, on-chain recompute, ev evidence tiers, rv undo journal. Nothing is claimed that wasn’t run. This is where most demos quietly die.
04
Mainnet / testnet / unaudited / in review — stated plainly, every surface. Then the next loop.
Proof the loop works: a personhood credential live on mainnet with a published SDK, a facilitator on mainnet, a spec extension with independent reproductions, text in a published IETF draft, and thirteen merges into other people’s repos.
What I actually ship with
The short version
Designing, building and shipping AI products and on-chain systems end to end: Proof of Real, the sui-x402 facilitator, x402-sui-stack (Sui Overflow 2026), the x402 charging agent, the rv / ev / xv accountability specs, a local-first desktop agent, and consumer apps people can buy.
Migrated Nodle’s L1 bridge off the deprecated ZKsync Mailbox to Bridgehub (Solidity; commits landed with authorship preserved after validation against live-mainnet state). Shipped release provenance for OpenMed: Sigstore-signed wheels + SLSA attestations, a release run-ledger binding artifacts to gate reports, and a reproducible rollback decision. Filed and reviewed fixes in Mysten’s Sui SDKs (a cached-price payment race in Walrus, a silent setSender() no-op behind Seal). Independent source-level audits of major Sui DeFi protocols, including a responsible disclosure of a latent oracle-quoter issue.
Two decades of hands-on, demanding work — reliability, problem-solving under pressure, seeing jobs through to the end. The same persistence now goes into software. The hours are similar; the splinters are fewer.