Now

North star

Build reliable systems that help people move faster onchain and with AI by tightening prompts, context windows, token usage, and agent reliability instead of chasing hype.

Current focus

Current loop: ship small AI apps, get usage signal, then tighten prompts, context strategy, token usage, and reliability.

• Excuse Me: shipped and now tightening tone controls, edge-case handling, and repeat usage loops.
• RoastMaster: improving first-run UX, output quality, and distribution loops that feel natural.
• AI systems work: making prompt templates, token budgets, and context assembly more reliable under messy real input.
• Polygon: shipping user-impactful work across payments, bridging, and staking with stronger end-to-end confidence and observability.
• Public sharing: writing practical AI notes on what failed and what actually held up.
• Operator tools: hacking on CLIs and TUIs.

Not now (on purpose)

• Chasing vanity posting volume over useful signal.
• Turning this site into a gimmicky personal dashboard.
• Over-worrying that some goals are paused while I double down on AI learning and agent orchestration; short-term sacrifice is worth the long-term leverage.

Open questions & asks

• I am looking for a small, high-signal builder community to learn and ship with — DM me on X if you are building one or want to join one.
• If you are building practical agent systems (especially in engineering or ops), I am up for comparing notes.
• Open to intros to thoughtful builders in onchain product and AI tooling.