- How I Keep Speed, Quality and Intent with AI Agents
I build NeuroTerm, an embedded terminal, alone with Claude Code and Codex. Written intents and a two-week release train keep speed, quality and intent.
- Why I Started Building NeuroTerm
My agents could not see the board's serial console, so I carried the logs. NeuroTerm takes over that job and keeps up as the technology changes.
- Making What I Don't Fully Understand
I don't fully understand the products I make. We always used things we couldn't see into, for reasons we trusted. As those reasons fade, what remains?
- harness-maker: Intent-Based Development
I write down my product's job and principles, then return to them each time I choose, specify and judge a feature. Notes from my tennis product, Tuned.
- LLM PCB Routing: What I Learned From a Human's Fixes
I handed a 96% auto-routed PCB to a contract engineer. He redrew finished routes, not only the last 4%. I kept that difference as 16 rules for my agent.
- harness-maker: Side vs Production
I run side and production projects together, and they need different workflows. Most harnesses are fixed, so I built a tool that makes one per project.
- Tuned: What I Don't Hand Off to My AI Team
An AI team writes most of my first drafts. The parts I must know and the parts I decide, I still understand to the end. Read wide, write in one lane.
- Tuned: The Two Seats Left to People
I re-sorted the twenty-eight disciplines a hardware product needs. Twenty end as files, so the harness took them. Six stop at a doorway. Two stay human.
- Tuned: A Roadmap That Marks Where to Quit First
A solo founder's roadmap is closer to a map of kill-gates than a schedule. Cheap, decisive checkpoints before costly stages, and permission to quit early.
- Tuned: The Co-Founder That Built Too Fast
The harness made building so cheap I built most of the product before checking anyone would pay. A good co-founder's risky side, a solo founder's trap.
- Tuned: What I Am Making, and Why I Make It Alone
A tennis player and 18-year embedded engineer builds an impact sensor alone, with an LLM as the team. Page one of a productization diary, kept either way.
- How Long Does Harness Engineering Have?
Everything useful in a harness ends up absorbed into vanilla. What survives is what the vendor structurally will not or cannot build. Notes on the shift.
- What loop would keep my software improving with LLMs?
A hypothesis I am still sorting out: whether I can wire my own project to LLMs and turn it into a loop that keeps improving itself.
- AI-Written Firmware: What to Hand Over, Where to Stop It
3-Tier Trust sorts PRs, then verification gates (spec, TDD, RAG, HIL, deterministic analyzer, more) take them to merge. Each catches its own defect class.
- AI-Written Firmware: What Breaks Outside the Function
AI-written firmware breaks where a single function can't show it. I found 3 patterns visible in the diff and 3 that only surface after 49 days.
- AI-Written Firmware: What I Didn't Tell It
A junior wrote a logging daemon with an AI tool. It passed my review, ran fine, and later wore out an eMMC. The risk was in what the prompt left out.
- AI Agents in Embedded: Opening the Full Pipeline
I open one AI-agent-built i.MX8M Plus sensor driver project, folder by folder: specs, property tests, review agents, and feedback loops.
- AI Agents in Embedded: Learning Context Engineering
AI agents wrote good Yocto recipes but silently corrupted sensor data and misconfigured SPI registers. What fixed it was the context I gave them.
- AI Agents in Embedded: The Spec-Driven TDD We Put Off
Embedded teams skipped TDD for its cost. With AI agents I watched that math change. Notes on testing a Yocto IMU driver from a spec, with JSON feedback.
- Hardware Docs for LLMs: Why I Went Back to grep
RAG cut the cross-references that make hardware docs useful. After wrestling with a pipeline, I went back to handing the LLM my SVD and PDF as file tools.