← Essays /Post · 13 of 28 · Making

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.

·3 min read · · · #build-in-public#solo-founder#ai-agents#embedded#hardware
Tuned: What I Am Making, and Why I Make It Alone
On this page

I play tennis. And for 18 years I have worked as an embedded engineer, building other people’s products.

There is a small sensor I have wanted for a few years, one that tells you where the ball hit the racket. The idea was never what stopped me. On my own, I could not carry it all the way to a product.

One person for firmware, one for the app, one for signal processing and ML. Then industrial design, marketing and certification on top of that. Even a palm-sized piece of hardware needs all of those roles at once, and no single person is good at all of them. So the answer was always one of two things: build a team, or drop it.

The moment it changed

This year, writing code alongside an LLM, I watched that assumption give way. It was not vague optimism. It came from running the firmware and the app myself, alone. A large share of those roles now fits in one pair of hands.

So I am placing a bet. Can a solo engineer carry a real hardware product all the way to launch? This series is the record of that bet.

It is not a finished conclusion. The bet is still open, and I might lose it. Either way I will write it down without pretending I won. To be honest, this is less something for readers than a productization diary I keep for the version of me who comes back to it later.

What Tuned is

The product is Tuned. Its sensor, Tuned Dot, reads where on the racket face the ball landed and gives you feedback.

It did not start with market research. It started with my own racket. I began because I wanted one. Whenever I make a product decision, being user number one weighs more than I expected it to.

How it picks up that location, the method underneath, stays out of this series. That part is the core of the product. Why I keep it closed, I explain at the end.

What I have today

I want to be plain about where things stand. There is no hardware yet that I would call a product. No custom board, no enclosure, no finished form factor.

What exists is two things. One is an app that holds the experience I want. The other is a firmware prototype running on a Nordic development kit.

Building the app first was deliberate. Before putting money and time into hardware, I wanted to hold the thing I was making. Once I could see on a screen where the ball hit, the product became clearer to me. That decision, what to build first, is itself the subject of later chapters.

So this is a different kind of thing from a “we’re starting a startup” announcement. A working app and a validated firmware path are already in hand. But the product I can actually sell is still zero. Turning that zero into one is what this series is about.

The order ahead, and what I won’t show

The rough order goes like this: the harness that turned one person into a team, scoping a roadmap a single person can finish, ordering hardware, a solo dev routine, testing with no QA, pricing, marketing, funding, certification.

At each step I write down one thing: where the tools wrote the code, and where the judgment was mine. The hardware bet, the architecture taste, what I refuse to give up, when to kill a direction. That judgment is what this diary is for.

The disclosure rule is short too. I write down the methods and the judgment. I keep the real numbers closed, meaning price, cost and funding, along with the sensing internals. It is a product I mean to sell. That boundary is what this series promises.

A year from now I want to read this back and see why I decided what I decided. Whatever the bet returns, the record stays.

ecro
Written by ecro

Created an LLM benchmark for firmware. EmbedEval →

Building a project-shaped agent harness for Claude Code, Cursor, and Codex. harness-maker →

Building a serial and SSH terminal where coding agents work on the board beside you. NeuroTerm →

Comments

Loading comments...