A working system, not a team to manage

I work spec-driven: the plan, the interfaces and the failure cases are settled before implementation starts. That makes the work separable, so I run coding agents against those plans in parallel rather than typing every line myself. What does not get delegated is the judgement, and that is the part a fixed price rests on: what to build and what to buy, where the risk actually sits, what gets reviewed before it ships, and when the scope has quietly moved. It is how a build an agency would staff with a small team runs here as one engagement, and why I quote a price for the outcome rather than a rate for my time.

Most connected products do not need custom silicon. On the GYMRACE scoring stations I chose a Raspberry Pi, an off-the-shelf LiDAR module with a plain serial interface, and 3D-printed mounts, because a part you can buy again in two years is worth more than a board that is five percent tidier. Custom hardware is a decision with a long tail: respins, lead times, certification, and a replacement problem when a component goes end-of-life. I design around that by default, and go custom when nothing off the shelf will do the job. Two products of mine have been through a full radio sweep for CE, which is mostly how I know when you can avoid one.

1. 01

   Scoping call

   You describe the problem, I ask questions

2. 02

   Proposal

   Fixed price, clear deliverables, timeline

3. 03

   Build

   Regular updates, no surprises

4. 04

   Handover

   Working system, documentation, source code