Software thatproves it works.

Davtar builds Alan, an AI software engineer. He plans, builds and tests real apps, and hands them over only when the tests pass.

A typical session, shown in miniature.

The problem

Writing code is no longer the hard part. Knowing that it works is. Most AI tools stop at code that looks right. Alan keeps going until he can show you it runs.

Three parts, each holding up the others.

  1. 1

    Plan

    Alan turns a plain-English request into concrete steps and shows them to you before he writes a line.

  2. 2

    Build

    He writes the code and installs what it needs inside a private machine made for your project.

  3. 3

    Prove

    He writes tests and runs them. If one fails, he fixes it. Nothing ships until they pass.

Meet Alan.

Describe what you need. Watch him plan it, build it and test it. When he's done, he tells you exactly what he added and what he did, and gives you a link to the running app.

See what Alan can do
✓
Your app is liveAll tests passing

Your HVAC dispatch board is live. You can log service calls, assign technicians and drag jobs between columns.

What I added

  • A board with Scheduled, On site and Complete columns
  • A form to log a new service call
  • Technician assignment on every job

What I did

  1. Built the API for jobs and technicians
  2. Designed the drag-and-drop board
  3. Wrote 4 tests, fixed one failure, all passing

Research

All research

We're hiring the first team.

Engineers and designers who want to build the software engineer they'd trust with their own work.

See open roles