This repository contains a specialized AI-assisted IDE for building applications whose front ends follow Generative UX architecture.
The IDE itself is not generative UX.
It is a stable configuration and compile environment whose sole purpose is to:
- help builders define app structure, data, and rules
- compile those definitions into applications
- ensure that the generated app composes its interface at runtime from user intent
Applications generated by this IDE share a common runtime model:
- no predefined screens or flows
- interface composition occurs at runtime
- user intent and uncertainty are first-class state
- UI is assembled from constrained primitives
- the system can explain why interface elements exist
- behavior is deterministic given the same inputs
These behaviors exist only in the generated application, not in the IDE.
This is how builders create and iterate apps:
builder intent → app configuration → compile → run → feedback → safe edits
The IDE:
- captures builder intent
- produces an explicit app blueprint
- compiles readable, inspectable code
- validates changes before applying them
See:
This is how the generated application behaves for end users:
user intent → context → composition plan → deterministic render → instrumentation
See:
Authoritative IDE documentation lives in docs/ide:
00-vision.md– why this IDE exists01-loop.md– Generative UX runtime (generated apps)02-architecture.md– system boundaries03-builder-loop.md– configuration & compile loop05-debug.md– runtime failure modes and debugging
docs/examples contains demonstrations and historical context.
These show what the IDE can produce, not how the IDE works.
Early, architectural, and intentionally constrained.
The focus is correctness, determinism, and explainability — not novelty or interface experimentation.
If you want to build applications with Generative UX front ends,
start in docs/ide.