Things we wished existed. So we're building them.

Not a studio, not a startup. A small place where we make software you can talk to, swipe through, or take a personality test on, learn whatever it takes, and write down what we learned. Sometimes the agents write it down for us.

What we're buildingRead the blog

What we're building so far

More on the way. The list is short. It won't stay that way.

Yapp macOS

Tap a key, talk, and your Mac gets on with it while you're still talking. Speech stays on your machine, and a small decision model turns each phrase into an action. Anything that could do harm waits for your yes.

buildingCC BY 4.0SourcePosts

No public build yet. It runs from source today; a signed build is on the list.

What Should We Watch iOS · Android

The question everyone asks at 9pm, answered before 9:20. Pick a mood, get ten films that are actually streaming tonight, swipe to decide.

on the App StoreProprietarywsww.appPosts

Android in testing

Spark Web

A personality test that makes up the questions as it goes. Each one is written from your last answer, the next one is already waiting because it guessed what you'd pick, and at the end it tells you which kind of teammate you are.

liveAll rights reservedSourcePosts

Try itV0.5This release introduces Ception’s hackathon-first personality engine, designed to help builders, founders, and teams understand how they operate under pressure — not just who they are.Release notes

Personal Portfolio Web

A new way to run a recruiter screen. Your own agent takes the call on your own site, answers from a record you control, scrolls to the evidence, and hands the recruiter notes. No calendar.

very early previewPrivatePosts

Read the postEarly preview

What's new

All posts →
  1. What Should We WatchNight Shift: an on-call engineer for a team of one

    When a production alert fires, Owl, a Claude Code agent, starts in the cloud within a second, works the runbook against live telemetry, and sends my phone a diagnosis with options. It can look at everything and change nothing. The design, the guardrails, and how to build your own.

  2. YappUse your Mac without touching the keyboard. Is this a better harness for LLMs?

    Yapp drives a Mac from your voice with a decision model that only picks from options it is shown. What that made easy, what it made hard, what 32 tasks on a real Mac say, and why an LLM might want the same loop.

  3. Personal PortfolioScreening rounds are changed forever.

    The first-round call is gone. Your own agent takes the screen, on your own site, from a record you control, and the recruiter walks away with notes and evidence. The early preview is live.

How this works

Small pieces, one job each.
Every module gets run on its own before it's allowed to talk to the next one.
Design first, code second.
Visuals are argued out in Claude Design and then ported. Nobody improvises a button in a PR.
Boring CI, on purpose.
Lint, types and tests on every pull request. feature/* → dev → main. Green or it doesn't go in.
Credit is the currency.
Yapp and the brand are CC BY 4.0: use them, just say where they came from. What Should We Watch is ours; read it, learn from it, ask before you ship it.

Want to collaborate?

Happy to. Open an issue, send a PR, or start a discussion. If you've got a thing you wish existed and it fits here, we'd like to hear about it.

Start a discussion