The maker loop: requirements, build, measure, repeat
Hardware, firmware, software and customers all respond to the same loop. A look at how I take an idea from first sketch to working prototype — and why the loop matters more than any tool.
Looking back across the projects on this site — SolveBlock, Dictyon-net, AffRank.ai, two decades of infrastructure and automation work — the tools changed constantly. The loop didn't. I'd describe it like this:
Requirements → Build → Measure → Repeat.
It sounds obvious. Doing it consistently is the hard part.
Requirements: write the one-pager
Every project starts with a single page:
- Who is it for?
- What problem does it solve, in their words?
- What must be true for it to succeed? (For SolveBlock: the external timer stays the source of truth.)
- What's explicitly out of scope for this version?
One page forces choices. If it doesn't fit on a page, it isn't understood yet.
Build: the cheapest thing that answers the next question
Each build exists to answer a question, and the best build is the cheapest one that answers it:
| Question | Cheapest answer |
|---|---|
| Can the ESP32 read the timer's data? | Dev board + jumper wires |
| Is the display readable across a table? | The display module taped to cardboard |
| Does the enclosure feel right in a hand? | A rough 3D print with no electronics |
| Does the board work? | A first PCB revision with lots of test points |
| Will people actually use it? | Put a prototype in their hands |
Notice that a custom PCB is well down the list. You earn the right to make a board by answering the cheaper questions first.
Measure: decide what "better" means before you look
Measurements only help if you know what you're measuring for. For each iteration I note:
- What I expected to happen.
- What actually happened, with numbers where possible.
- What I'll change next.
That's true for bring-up logs, power budgets, mesh routing tests, SEO changes and onboarding flows alike. The format is the same because the loop is the same.
Repeat: make iteration cheap
The speed of the loop is what matters, so I invest in anything that makes iteration cheaper:
- Parametric enclosures in OpenSCAD, so a dimension change is one line.
- Reusable circuit blocks, so a new board starts from proven pieces.
- Automated checks and tests, often AI-assisted, so reviews take minutes instead of hours.
- Version control for everything — firmware, PCB files, enclosure code and docs.
Where AI fits in the loop
AI tools accelerate every stage: drafting requirements, generating test cases, checking schematics, writing documentation. They make the loop faster. They don't replace the measure step. Datasheets, physical hardware and real users remain the final reviewers.
Building with people you care about
Some of the best iterations happen when the person who has the problem is right there at the bench. SolveBlock exists because its co-creator is a speedcuber who could hold every prototype and say, immediately and honestly, "this is in the wrong place." That kind of feedback is worth more than any tool on this page.
The point
Tools come and go. I've moved from Cisco routers to Citrix farms to cloud APIs to KiCad and microcontrollers. What carried me through each change was the loop. If you're starting your first project — hardware or software — don't worry about picking the perfect tools. Write the one-pager, build the cheapest thing that answers your next question, measure it honestly, and go again.
That's the whole secret. The rest is practice.