Chalant

Every Mac ships with a strip of dead space around the notch. I wanted something living there. A small glass panel you could talk to or type at, that handles the little things without pulling you into another app: create a reminder, send a text, play music, run a focus timer.

Two decisions shaped everything that followed.

• —On-device, keyless. No accounts. No API keys. Nothing leaves the Mac. That constraint killed a dozen tempting features, and it’s the reason the product is trustable. You can adopt it in one download without handing anyone anything.
• —Quiet by default. Chalant waits. It doesn’t badge, nag, or interrupt. A tool that lives on your screen all day earns its place by being calm, not by demanding attention.

The bet: a utility this small, this private, and this polished would be worth shipping, and that one person could ship it.

This is the part I’d want a hiring manager to read.

I don’t write production code. I directed AI development tools through the entire build, and treated that like a real engineering process, not a magic trick:

• —Specs first. Every feature started as a written spec: what it does, what it never does, how it should feel, what “done” means.
• —Review everything. AI produced the code; I held the quality bar. If a build felt wrong (timing, animation, edge cases), it went back with notes, like design review for software.
• —Iterate relentlessly. 116 build iterations. 36 automated tests. Roughly 14,000 lines of native code produced across the process.
• —Protect the constraints. Every iteration got checked against the two founding decisions. Anything that needed a server, an account, or an interruption didn’t ship.

The interesting part was never the code. It was deciding what to build, what to cut, and how it should feel, then holding a bar until the build matched the spec. That’s the job.

A demo on your own laptop isn’t a product. Chalant is shipped the way real software is shipped:

• —Self-updating. It delivers its own signed updates. Users install once and stay current.
• —Distributed three ways. Direct download, Homebrew, and the landing page, meeting both regular users and terminal-native users where they are.
• —Light enough to forget. It idles at ~0.35% of one CPU core. An always-on utility has no right to be heavy.
• —Open at the edges. A localhost API lets other programs post to it, so Chalant can sit inside other people’s workflows, not just mine.
• —Launched properly. Landing page, positioning, and a launch video: the full go-to-market, owned end to end.

Three decisions that tell you more about how I work than any feature list: