Let me show
you around.
I’m Mylo. I built Mad Moon to help me move faster, spend less on AI, and ship work I can stand behind.
From the first idea to the final checks.
Good AI work takes
more than a good prompt.
Mad Moon is the development environment I’m building around AI agents: a way to define the work, give it useful context, and check what comes back.
I connect the tools, engineer the boundaries, and stay responsible for how the pieces fit together.
See what I built and what I integrateA little trip through my process.
Stop 1 of 5Make yourself at home.
Follow a cycle, or open something that catches your eye.
The integration is part of the engineering.
I use native capabilities where they fit, then add code and guidance for the coordination gaps: identity, selected configuration, owned changes, failure handling, and evidence.
Curiosity is good.
Receipts are better.
Here’s what supports the tour, and where the evidence stops.
Built and checked.
The September 11 verification pass used existing tests and bounded native observations. It covered workflow, navigation, identity, guarded changes, and capture boundaries.
210 test passes · 1 explicit native-check skip
Still a question.
Does this environment make my work faster, cheaper, or better overall? That needs a comparative evaluation. It hasn’t been measured yet.
See the proposed evaluation loopA research shelf, not a scoreboard.
Research explains why these choices are worth exploring. Its results belong to the systems and tasks it studied.
That’s my corner of the universe.
A growing environment for working with AI, built one deliberate decision at a time.