Start here

WornKeys is where I write up what I build and what I learn on the way: hands-on notes, labs you can run yourself, and a sober look at new tech. If you’re new, this is where I’d start.

A build you can run: AI Voiceover Studio

A self-hosted studio that turns a description of your video into a voiceover in your own voice, with consent and watermarking built into the code. Three parts, and the code is public on GitHub.

1. Your Voice, Your Script
   How voice cloning works, which open models you can build on, and running the studio in your own AWS account.
2. Inside the Studio
   From a video description to your voice: the pipeline, a script written for the ear, and two models sharing one GPU.
3. Four Layers Between You and a Surprise Bill
   What the deploy builds in AWS, who is allowed to press it, and what it really costs.

Two essays on AI and engineering

• AI Myths: What a Principal Engineer Actually Sees Day to Day
  Why “AI belongs everywhere” is wrong, why commit counts are a broken way to measure productivity, and why prompt precision saves real money.
• What Would It Take for You to Never Read Your Code Again?
  Will a specification go straight to working software, with the code in between optional? The two bets that would have to hold: cheaper tokens and better specs.

The sections

Field Notes

Notes from hands-on engineering: what I tried, what held up, and what didn’t.

Keystrokes

Labs you can run yourself, with the code, the real costs and the gotchas.

Signal

Tech, tools and trends that look promising, and whether they hold up.

Workbench

Apps I build and look after alongside the blog.

Behind the Keys

Who’s writing this, and why.

Keeping up

New posts appear in the RSS feed. Comments under each post run on GitHub Discussions, so you’ll need a GitHub account to join in. Anything else: see Contact.