~/ai

ai writes the boilerplate. we own the decisions.

Most of what ships from Project Irisyle passes through an AI coding assistant at some point — scaffolding, refactors, docs, first-pass debugging. We're not precious about where the keystrokes come from. We're precious about what ships: reviewed, tested, and owned by an actual person who can explain why it works.

what it's good for

scaffolding/

boilerplate, gone

New resource skeletons, config schemas, repetitive CRUD — the stuff that used to eat a weekend now takes an afternoon. That's more free releases, faster.

docs/

docs that actually get written

README, API reference, changelogs — the stuff that gets skipped under time pressure gets written, because a first draft costs minutes instead of hours.

refactors/

cleanup without the dread

Mechanical renames and restructuring across a whole codebase, reviewed in one pass instead of drip-fed over weeks of "we'll get to it."

debugging/

a second pair of eyes, on demand

A stuck bug gets a second opinion immediately instead of waiting on whoever's online in Discord.

what doesn't change

Every PR is read by a person before it merges — AI-authored or not.
Security-sensitive code (auth, permissions, payments) gets extra scrutiny, always.
Licensing and pricing decisions are made by us, not a suggested default.
If something ships broken, a person is accountable for the fix — not "the AI."

why there's a badge in the footer

The REAL 5 mark is self-reported, scored against the public rubric from realgoodai.org — it's not an independent audit, and we're not claiming it is. We picked it because the criteria are specific and public, so you can check our answer against the rubric yourself instead of taking our word for it.

view the rubric ↗