Software and hardware engineering

We build the whole thing.Or the next part of it.

Web and mobile apps, backends and APIs, AI and agent systems, embedded firmware, electronics, and physical hardware. One team across all of it.

Come to us with a blank page, a spec, a working prototype, or a product already serving customers. We pick it up wherever it is and take it as far as you want to go.

Things you can click on right now. Every one is live and in use, and together they’re the short version of what a small team can ship when the tooling stops getting in the way.

Rather than a list of services, here’s the work itself. Find the line that sounds like your project, and follow it to something we actually built.

Open questions we’re working on in the background. Some turn into products, some stay strange prototypes, and some fail in ways worth writing down.

  1. PCB CAD, Rebuilt for Touch “What if professional PCB design actually belonged on an iPad?” Integer-nanometer geometry, a hand-written maze router, and byte-preserving KiCad round-trip. Not desktop CAD squeezed onto a tablet. ● Live as PCB Canvas · pcbcanvas.dev Active
  2. Backbone “What if the model didn’t have to fit on one machine?” Distributed inference across mismatched hardware. Macs, graphics cards, old workstations, phones. Whatever compute is around becomes part of a larger system. Active
  3. CleverKeys “What if one person could ship a swipe keyboard that competes with the platform vendors’, and never phones home?” On-device gesture decoding under 200 ms, zero network permissions, public training pipeline, reproducible builds. Open source and installable today. ● Live · cleverkeys.app Shipped
  4. termux-tools “What if your phone were a real development machine, with agent and browser automation?” Claude Code, Bun, a de-telemetered browser, and browser automation running natively on stock Android. No root, no proot, one installer. ● Live · termux.party Shipped
  5. Mayhem Monkies “What if your QA department was software?” Specialist agents hunt UI problems, API failures, regressions, edge cases, and security issues. Every bug comes back as a fix, not a ticket. ● Live · mayhem.feedback Shipped
  6. SIDELOBE “What if a field computer cost what its parts cost?” A custom handheld field computer built from inexpensive microcontrollers, radios, sensors, displays, and modular hardware. In progress
  7. ACID CAT “What if the agent didn’t need the application’s permission?” A USB device that presents as an ordinary keyboard and mouse, so automation runs in any application on any operating system. Nothing to integrate. ● Pre-order · acidcat.dev Pre-order
  8. Currently Doing Something Inadvisable No further information. In progress

// For people building with AI

Built it with AI? Keep going.

Cursor, Claude Code, Lovable, Replit, v0, or whatever lands next month. Getting a long way with one of them is a genuine head start. Bring us what you’ve built and we’ll take it as far as you want to go.

  • The parts that were always going to be hard. Auth, payments, real data, load, the things that only show up once people are using it.
  • A prototype a whole team can build on. Same product, structured so more than one person can ship it at once.
  • Hardware too, if that’s where it’s heading. Firmware, electronics and physical devices come from the same team.

We work alongside you, at your pace, on what you’ve already got.

The minimum viable organization is shrinking.

Ambitious engineering used to require organizations, because organizations were how you aggregated specialized knowledge, labor, capital, and infrastructure. Those constraints are collapsing, and the only honest way to find the new boundary is to build things and push on it.

Read the manifesto →

Tell us what you’re building.

A straightforward build is normal work here. So is the cross-disciplinary thing nobody has quite built yet.

Both get the same team and the same attention. Come with a clear spec or a rough idea, something new or something already running, and we’ll tell you what it actually takes.

  1. Tell us what you’re building.

    Book 30 minutes, or write it out instead. Come with the project, not a pitch deck: what you’re making, where it is now, and what done looks like.

  2. Get a straight feasibility read.

    What’s actually hard, where the leverage is, and whether we’re the right people. Free, and we’ll say so if we’re not.

  3. Scope something with a fixed shape.

    A prototype, a full build, an investigation, an architecture, or the next phase of something already running. Bounded, priced up front, and ending in something that runs.