Build software with agents

Machina Practica is a system for building reliable software. It consists of a programmer's guide, skills to make the agent effective, and reusable software packages that ensure high quality results.

You'll learn how to direct the agent effectively. The agent gets instructions that match this way of working. Tested components mean reliable results, across a variety of projects and problem domains. Welcome to the future of software development.

Read the method

One system, three parts

A programmer's guide. Effective Agents teaches you to direct the agent effectively: define the result, delegate the work, and assess what it delivers. Read about Effective Agents.
Skills for the agent. Equip the agent to research, plan, build and verify software using the method you learn. How the skills work.
Reusable software packages. Build on tested components with consistent behavior across projects. The first packages handle test records and build identity. Available packages.

Reliability by design

The system sets clear engineering standards for every project:

Design away races. Make the order of state changes explicit and wait for the state you need.
Make CI 100% repeatable. Fix the environment and inputs so the same change produces the same build and test results.
Make screenshots deterministic. Control the clock, data, fonts and rendering environment. Expect zero pixel differences in that environment. Fix unexplained differences instead of hiding them with masks or fuzzy comparisons.

The guide teaches these standards, the skills apply them, and the components support them. Every failure gets investigated and fixed.

What you can use today

Eight experimental skills are available in source. Two packages are published as alpha releases. Effective Agents has a draft table of contents and first chapter. The companion book, Machina Practica, will walk through the skills and their goals.

The skills haven't yet been used to create a project, and the packages have no consumer adopters yet. Validation of the complete system in real projects is the next step. Start with how the method works, or read what we're trying to achieve.