01 · Mar–Sep 2026Live
Agentic Payments Research
Six months of primary research into how AI agents pay each other over crypto rails, and an honest account of what shipped.
- Primary-source research
- Obsidian vault
- Claude Code
- HTML Field Manual
- +2
Open to ecosystem & BD rolesSep 2026 · 16 projects · 6 live
I research where crypto and AI meet, then build prototypes to find out if I’m right.
Business development and ecosystem operator. Six months of primary research on agentic payments, tokenization, and what institutions actually do on Solana.
- TypeScript
- Next.js
- Supabase
- React
- Vanilla JavaScript
- Vercel
- Anthropic API
- Playwright
- shadcn/ui
- Tailwind
- Vercel Cron
- Vitest
- Apple Health
- CC BY
- Claude Code
- Claude Design
- Cloudflare Workers
- CSS
00By the numbers
- 16projects, all written up
- 3in production with real users
- $100M+partner volume closed at Wormhole
- 455research notes, Mar–Sep 2026
- 721commits on one six-week build
02Headliners
Four builds that test the thesis.
Each one written up as what I built, what I learned, and where it cut against my own view. All directed, none hand-coded.
02 · Apr 2026Prototype
x402 Verifier Facilitator
x402 facilitator prototype with an LLM-as-judge work-quality endpoint. Four testnet settlements; the verdict never got wired into /settle.
- TypeScript
- pnpm monorepo
- x402 (@x402/core, /express, /fetch, /evm)
- EVM testnet
- +3
03 · Sep 2026Live
Fieldnotes
Editorial reference site on agentic payments, Solana liquidity and institutional adoption, with a day-gated data refresh and review flags.
- React 19
- TypeScript
- Vinext
- Cloudflare Workers
- +3
04 · Jun 2026Demo
Proof Market Maker
An always-on market-making bot with inventory control, self-healing reconnects and a two-stage kill switch, plus an operator dashboard.
- TypeScript
- Node.js
- Proof.trade SDK
- Next.js (ops dashboard)
- +3
All 16 projects Shipped products, experiments, and the ones that stalled.
01Thesis
Priors, written down.
Three positions on where crypto and AI meet, starting with the essay I published in April. Each opens with the 30-second version and closes with what would make me wrong.
01 · Thesis23 Apr 2026
The Future of the Agentic Economy
A second economy is forming alongside the human one. It runs on agents, settles in stablecoins, and needs a verification layer nobody has built.
Agents will have an economy more diverse than the human one, made of specialists hiring specialists for cents at a time. Speed and cost are half the problem; the other half is proving the work was done. Without a verification layer that grades output and decides whether payment goes through, the whole thing runs on trust, and nothing runs on trust at fractions of a cent. Stablecoins and programmable rails are the only money that fits.
Read the essay02 · Thesis9 Sep 2026
Verification is the scarce resource
In agent-to-agent payments the money already moves. Nobody has solved how the paying agent knows the work was any good.
Agents can find each other (MCP, A2A) and pay each other (x402, USDC, smart wallets). The one layer with no production answer is judging whether the paid-for work was actually good. Payment rails earn on throughput, so none of them want to be the party that says "this doesn't settle." That leaves the verification seat open, and the party that decides whether to release payment holds the position of trust. I built a prototype to test this in April 2026, and the data I collected cut against my own model on volume and price per transaction. I still hold the thesis at medium conviction, and this essay says exactly what would move it.
Read the essay03 · Thesis9 Sep 2026
What mainstream adoption looks like next
Solana spent the last year winning the argument and losing the trade. Here is what I think happens after that.
As of 2026-09-03, SOL traded at $104 against a $293 all-time high, DeFi TVL had halved to $5.8B, and the market had stopped paying for throughput, even as every institutional milestone landed. I run four questions to keep this straight: is the tech getting better (yes: Alpenglow passed 2026-09-02, Firedancer is on mainnet), are institutions here (yes, see the tokenization essay), is anyone making money on-chain (apps yes, $257M in Q2 2026; holders less so), and does any of it accrue to SOL (unresolved). The next phase of adoption is not consumer apps. It is stablecoin settlement at payment networks and agentic payments, where Solana already carries roughly 65% of x402 volume. Whether that reprices the token is the open question, and I would rather say so than point at transaction counts.
Read the essay03How I work
Direct the work. Judge the outcome.
I don’t write code by hand. I give AI agents a spec, a plan and a definition of done, then check what comes back.
- 01
Spec
What is this for, who uses it, and what does “working” mean. Written before any code.
- 02
Plan
The spec broken into tasks small enough that an agent finishes one and I can verify it.
- 03
Build & judge
Agents build in parallel; a second agent verifies; I judge outcomes the way a founder checks a contractor.
Recent monthsFull timeline
- Sep 2026
Solana knowledge vault, Field Manual published, this site
Built a 74-note Solana knowledge base (2026-09-02 to 09-05) around four questions. Published the Agentic Payments Field Manual on 2026-09-06. Shipped this portfolio in three days with five parallel agents.
- Aug 2026
Marathon public demo; Fieldnotes refresh pipeline
Added a public demo mode to the marathon app (2026-08-24) so visitors can use it without an account. Designed the day-gated data refresh with human review flags that Fieldnotes runs on.
- Jul 2026
Marathon training app; VC thesis vault
Started the marathon app on 2026-07-01: Supabase with row-level security, Whoop and Strava sync, an hourly cron pinned to 6am local time. Began the VC and thesis vault where the sector theses on this site originate.