One screen becomes the entire feature.
Persistent state, derived state, operations, events and UI concerns pile into one class.
MFR is a feature-oriented reactive architecture for applications with shared, interconnected state across multiple screens and complex workflows.
The question that eventually became the architecture.
Feature-lived state stays with the feature.
Don't manually synchronize what can be derived.
MFR moves the ownership boundary above individual screens and makes state derivation a declarative relationship.
Persistent state, derived state, operations, events and UI concerns pile into one class.
Shared state is duplicated, re-fetched or passed around just because navigation changed.
Multiple mutable values need imperative updates to keep different representations in agreement.
MFR is Model · Formula · Reflector. The supporting architecture keeps operations, events and data concerns separated around those three roles.
Model — FluxDeck: owns feature-lived state and acts as the source of truth.
Formula: declaratively derives state from existing state.
Reflector: reflects feature state to the UI and owns genuinely screen-specific ephemeral state.
Coordinates operations and UseCases without turning the state owners into operation hubs.
Supporting result-handling infrastructure for standardized results and transient events.
Keeps one-time UI effects separate from persistent feature state.
Define the operation and data boundaries around the feature.
The Continuum is the feature's continuous memory.
Destroy state. Recreate state. Fetch again. Pass context around. That's what fragmented ownership tends to create.
The feature keeps its state for the workflow's lifetime while each screen reflects what it needs.
If you've read Solutions come from surviving, you already know the origin: a real Android project, a ViewModel that quietly became a God Class, and the repeated question of who should own the feature.
No paywall on the ideas — read the public work on GitHub ↗This is for the part after you've read the free docs and thought, “Okay, I get it — now I actually have to build all this.” The architecture stays free. The prepared setup saves you the time of assembling it yourself.
Prepared structure + AI workflow + migration resources.
Get the complete setup ↗You can still learn and implement MFR yourself for free.
The prepared setup: the structure template, AI implementation resources, migration resources and supporting checklists that save time from assembling them yourself.
Yes. That is intentional. The public documentation explains MFR so you can learn it and implement it yourself.
No. AI can accelerate implementation, but you should monitor it, check its architectural decisions and make sure it follows the MFR guidelines.
Make a copy of the project first, migrate and verify there, then apply the validated changes to your main project.
I'm Vedant R.J. Chourey, an Android Developer working with Kotlin and Jetpack Compose. I enjoy building products, thinking about architecture, and figuring out how software behaves when a project grows beyond the simple version we started with.
MFR came from real-world Android development — a ViewModel that quietly became a God Class, and the question I kept asking: “Who actually owns this feature?”
Outside software, I'm also a writer. I've written a fiction saga and published my novels through Amazon Kindle.