Bohea

Notes

Short engineering essays on building at the seam of software and machines: the design decisions, the failure modes, and the patterns I reuse. Every one of them either comes with something you can run, or is a post-mortem of something I actually fixed. No trend pieces, no founder reflections, no stack comparisons without measurements — if a note doesn't clear that bar I don't publish it.

6 Aug 2026
Industrial · Trust

You can't see a refactor. So try to break the machine.

You paid for a refactor and nothing looks different — same paths, same screens. My answer: don't review the code. Take a break-it checklist to the machine. If you can't make it fail unsafely, you've seen what you bought.

15 Jul 2026
Safety-Critical · Motion Control

Industrial reliability: change the denominator

You can't count the bugs that will hurt you — they're the ones you couldn't imagine. So stop counting bugs and count consequence classes. How to make a machine safe when you already know your software has bugs you haven't found.

3 Aug 2026
AI × Safety-Critical

I use AI heavily. The machine still can't hurt you.

Every time I ask AI to review its own code, it finds new bugs. Not using it means being generationally outpaced. On software that moves a physical machine, how do you weigh that? Here's my answer.

15 Jul 2026
AI Agents · Governance

Governing AI agents that touch money: the agent proposes, it never executes

Every write with a material consequence should leave the model's reach and pass an authorization layer the agent cannot author or bypass. Governance is architecture, not a careful prompt — with an escrow flow as the worked example.

15 Jul 2026
Self-Hosted · Privacy

A private AI assistant that never leaks: designing the outbound boundary

"Your data never leaves your machine" is only real if you can point at the single place it does leave — and prove nothing else can. Trust is won by drawing the boundary precisely, not by promising more.

16 Jul 2026
How I Work

Why I lead with a runnable demo

A quote is cheap and a claim is easy. Handing you something that runs — on mock data, before you spend a credential — moves the risk off you and onto me, where it belongs.

3 Aug 2026
How I Work

One email per milestone

You want "an HMI," and that's the whole spec. I split the work into milestones so it can't spiral — and run each one with a single email. From your side, one big risk becomes small stages you just watch complete.

8 Aug 2026
How I Work · AI & Judgment

AI extrapolates fluency. I extrapolate emotion.

"AI does the technical part, the human does the human part" is backwards. The two fail in opposite directions — and quality comes from the correcting pair, not from either side. The review protocol I actually run on client mail.

16 Jul 2026
AI & Judgment

Where value goes when building is free

When a cost goes to zero, value doesn't vanish — it moves to the new bottleneck: judgment, verification, and a person who can be held accountable for a machine that can hurt someone. Something a model structurally can't supply.

New notes land here first. To get them without depending on any platform: subscribe to the feed.

← Back home