Not a fix. An architecture.

A way to structure firmware so that the ISR boundary is a contract, not an assumption. So that component communication is typed and visible, not implicit and buried. So that the next engineer who touches the code doesn't have to reverse-engineer what the first one was thinking.

I sketched it. Then I closed the notebook.

The idea was sound. The timing was wrong.

Building a framework — a real one, not a prototype — means writing it across multiple RTOS targets. Writing the hardware abstraction layer. Writing the tooling. Writing the documentation that makes it usable by someone who isn't you.

That's a two-year project for a small team.

For one engineer with a full-time job and client work: a permanent backlog item.

So I kept doing what everyone does. Rebuilt the same infrastructure on every new project. Watched the same class of bugs appear on schedule. Wrote the same post-mortems with slightly different serial numbers.

What changed isn't the problem. The problem is exactly what it was in 2021.

What changed is the cost of building it.

AI-assisted development didn't give me the architecture — I've had that since that notebook. What it gave me was the capacity to execute it without a team. To move from "this should exist" to "we are building this" without the two-year runway I didn't have five years ago.

That's not a small thing.

For a certain class of problem — the kind where deep domain knowledge is the real bottleneck, not raw coding hours — the build-to-quality cost just dropped significantly.

Three months ago I stopped waiting.

The architecture is done. The contracts are written. The first code is compiling.

We're building the layer above the RTOS that nobody has built properly yet.

More soon.

If you've had an idea you knew was right but couldn't execute alone — what was the constraint that held you back?

— Ritesh ritzylab.com

#EmbeddedSystems #IoT #FirmwareDevelopment #OpenSource #EmbeddedEngineering