about
Engineer. Builder. Systems + AI.
I work full-stack on payments, bridging, and staking at Polygon, focused on shipping systems that improve speed and confidence through testing, reliability, and clear docs. Outside work, I build with AI and use personal projects to sharpen agent orchestration in real-world workflows.
last updated: 2026-05-18
// Direction
north star
Build systems that help people move faster with confidence.
at work
make complex systems reliable, testable, and operable.
outside work
keep refining practical AI and agent orchestration through fast iteration and shipping.
// Current loop
Current loop: bigger scope at work, sharper public tooling outside work, and cleaner systems at home.
Polygon
moving into Open Money Stack while still supporting core application work when needed. Same priorities as ever: reliability, testing, docs, and operability.
CLI tools
tfl-cli, companies-house-cli, rail-cli, fuel-cli, and parliament-cli — a growing suite. I am tightening the agent-first pass across all of them: predictable output, cleaner JSON, better landing pages, and better surrounding skills.
OpenClaw stability
cleaning up the upgrade churn so the stack is boring again: fewer false alarms, healthier crons, clearer memory behaviour, and less dead operational baggage.
Open Markets mini-series
shipped a three-part AI-assisted audio series on DEXs. It is a useful portfolio piece for voice, scripting, and shipping something weird but real.
Family ops
putting more structure around shared planning, nutrition, and training so personal life is not running on loose notes and good intentions.
// How I work
- Public by default. If it can help others and is not sensitive, it should be in-channel, not buried in DMs.
- Direct and clear. Cut fluff, get to the ask, and respect people's time.
- Async first for documentation and reuse. Sync when collaboration genuinely benefits.
- Meetings need value. I should be giving value or getting value I cannot get from notes, transcripts, or recordings.
- Feedback should be direct and honest: understand the why, then help unblock.
- Priorities: ownership and speed.
// Open questions / asks
- Always interested in comparing notes with people building practical agent systems, especially CLI/TUI tools or operator workflows that have to work in the mess of real life.
- Part of a small, high-signal builder circle now. If you ship, share the rough edges, and want the occasional useful call instead of endless noise, reach out and I can see whether an invite makes sense.
// Stack I actually use
The stack changes, but the bias is stable: fast local loops, durable context, cheap routing, and enough observability to know when the agents are bullshitting.
AI + model stack
- Coding
- OpenAI Codex 5.5 as the main heavy-lift coding model
- Planning/review
- Claude Sonnet/Opus when I want a second opinion or sharper critique
- Provider routing
- Hermes routing across OpenAI, Claude, Gemini, OpenRouter, and local-ish tools where useful
Agent operating system
- Harness
- Hermes Agent. OpenClaw did the first messy tour of duty; Hermes is the current system.
- Assistant shape
- Giles as the main operator, with thinner specialist profiles only when they earn their keep. No swarm theatre for the sake of it.
- Memory
- Hermes memories for durable facts, session search for recall, and markdown ledgers for curated context that should survive model churn.
- Skills
- Reusable Hermes skills for coding, research, Gmail, UK public-data CLIs, tracking, meal planning, writing, and project workflows.
Agent observability + tools
- Hermes Console
- Read-only local UI for Hermes runtime health, sessions, cron, skills, memory, files, config, logs, and usage. Built because digging through five places gets old fast.
- llm-usage
- Small CLI for tracking token usage across Claude Code, Codex, and OpenRouter. Boring visibility, extremely useful.
- Local CLIs
- gog for Google Workspace, bird for read-only X/Twitter, plus UK-data CLIs for TfL, rail, Companies House, Parliament, and similar agent-friendly lookups.
Dev environment
- Editor/IDE
- VS Code, with terminal-native agents doing a lot of the heavy lifting now
- Terminal
- Ghostty on macOS, Kitty on Linux
- Notes
- Markdown files in git + wiki for durable knowledge; Notesnook for personal notes
// Not now
- Dragging weak projects along just because time is already sunk. If the signal is bad, I would rather cut scope and move on.
- Building dashboards or automation that look clever but do not change behaviour.
- Trying to keep every work, builder, and personal lane equally hot at the same time.