Small Intelligences
Active development

Mister M

A private, local-first sensei that reconnects everyday practice to the larger Path a person chose.

Category: Small Intelligences
Platform: iPhone · iOS 18+
Status: Active development

What if guidance protected meaning instead of measuring productivity?

Structured Path memory, voluntary conversation, practice context, reflection, and guarded on-device generation turn a moment of drift into one honest next action.

User value

Return to what the practice means.

Mister M helps when the next action is known but its connection to the larger direction has gone quiet.

Recover the larger reason

Connect today's difficulty to the goal, transformation, practice, and obstacles stored for the active Path.

Receive specific guidance

Use reviewed local context to shape a response instead of treating every moment as generic motivation.

Return without surveillance

Come voluntarily, keep daily conversation on device, and continue without streaks, shame, or automatic cloud fallback.

Mister M iPhone screen showing a Writing Path, local conversation boundary, and today's practice with its meaning and a common obstacle. — **The active Path**Simulator capture using fictional Writing Path data stored locally for the demonstration.

Features

A focused local experience for returning.

The current build connects Path setup, conversation, practice, reflection, memory, and plan review.

01
Reviewed Path setup
Define the Path, larger goal, transformation, reason, practice, friction, recurring excuse, and guidance style across eight saved steps.
02
Path-grounded conversation
Store messages locally and prepare on-device replies from a compact packet of relevant Path context.
03
Today's Practice
See one practice beside why it matters and the local obstacle that may be making it harder.
04
Reflection and memory review
Capture friction, inspect suggested memory changes, and decide what becomes durable Path context.
05
Local plan context
Inspect an accepted plan, its phases, evolution notes, and narrow adjustments before applying them.
06
Memory controls
Edit Paths, delete conversations, and remove supported categories of locally stored Path memory.

How it works

Name the Path. Return deliberately.

The product moves from durable meaning to a small action without turning the process into a score.

01
Choose the Path
Name the direction and the larger outcome that makes it worth pursuing.
02
Describe the practice
Record the transformation, ordinary work, recurring friction, optional reason, and useful guidance style.
03
Review local memory
Edit the complete draft before creating the Path on the device.
04
Name the present moment
Write what feels difficult or begin from an editable prompt such as I need the bigger picture.
05
Receive guarded guidance
Retrieve relevant context, generate on device when available, validate the candidate, and persist only an accepted response.
06
Return to practice
Leave with a clearer relationship to one honest next action rather than another dashboard to maintain.

Current interface

Meaning is structured before it is generated.

Simulator captures from a clean local workflow using fictional sample content.

Mister M First Path Setup asking which Path the user is choosing and presenting several example directions.

Choose a Path

The first setup step names the direction the user wants to return to when motivation fades.

Mister M review screen explaining the local draft and showing fictional Writing Path details before creation.

Review before creating

Setup answers become an editable local draft, with no model or cloud call creating the user's Path.

Mister M local conversation composer with no-cloud disclosure, editable prompt starters, and a message field.

Keep conversation local

Prompt starters remain editable, user messages stay on device, and unavailable generation does not trigger a cloud call.

Product concept

The central object is a Path, not a task.

Mister M is designed for moments when a person knows what to do but has lost emotional contact with why the repeated action matters.

Conviction before compliance

The product supports voluntary return rather than measuring obedience through streaks, reminders, pressure, or shame.

Meaning has structure

A Path connects the goal, desired transformation, practice, friction, reflection, and preferred form of guidance.

The user owns the memory

The application stores reviewed facts locally. Generated suggestions do not become durable truth without an explicit review path.

Architecture

The app remembers. The model proposes language.

Local structured memory, deterministic policy, and guarded persistence surround the on-device generation step.

01Locally persisted user message
02Safety precheck
03Relevant Path context retrieval
04Mode selection and structured prompt
05Foundation Models availability check
06On-device candidate response
07Safety and quality validation
08Guarded local persistence

Structured local memory

SwiftData stores Paths, practice, friction, conversations, reflections, plans, preferences, permissions, and related state behind repository protocols.

Deterministic authority

Code selects context, applies safety and quality policy, controls memory changes, and decides whether generated output may be shown or stored.

Honest unavailability

If on-device generation is unsupported or fails validation, the user's message remains local and no synthetic or cloud reply replaces it.

Cloud planning stays separate

Optional planning has minimized draft contracts and validation boundaries, but the current runtime has no configured transport or live provider call.

Technology

Local memory around bounded generation.

Application

Swift
SwiftUI
iPhone-first
iOS 18 deployment target

Persistence

SwiftData
16 stored model types
Repository protocols
Apple Keychain foundations

Intelligence

Apple Foundation Models
On-device when available
Structured prompt packets
No cloud chat fallback

Policy

Deterministic context selection
Safety gates
Quality validation
Guarded persistence

Validation

XCTest
1,056 passing tests
4 skipped tests
Manual testing ledger

Development

Core local flows are implemented.

The iPhone application, local persistence, Path experience, and guarded Foundation Models pipeline are working and heavily tested. Release distribution, live-model quality, real-device persistence, accessibility, and the manual ledger remain unverified.

Stage: Active development
Automated suite: 1,056 passed
Daily cloud fallback: None
Manual ledger: Not started

Implemented

Saved eight-step onboarding, editable local review, and Path creation
SwiftData persistence for Path memory, conversation, practice, reflection, plan, onboarding, permission, and cloud-request records
On-device Foundation Models adapter with context retrieval, response modes, safety checks, structured prompts, validation, and guarded persistence
Today's Practice, practice meaning, reflection capture, reviewed memory suggestions, continuity without streaks, and memory deletion controls
Local plan overview, phase detail, reviewed adjustments, evolution notes, and plan context in daily guidance
Keychain secret-storage foundations plus hosted and BYOK draft boundaries that remain unavailable in the runtime

Release focus

Reconcile deployment and capability gating, decide the first-release scope, complete supported-device and live-model testing, validate persistence, deletion, and accessibility manually, then finish public privacy, support, and distribution materials.

Mister M lab notes

The project, over time.

Notes about the thinking, architecture, development, and decisions behind this work.

Return to what the practice means.

Recover the larger reason

Receive specific guidance

Return without surveillance

A focused local experience for returning.

Reviewed Path setup

Path-grounded conversation

Today's Practice

Reflection and memory review

Local plan context

Memory controls

Name the Path. Return deliberately.

Choose the Path

Describe the practice

Review local memory

Name the present moment

Receive guarded guidance

Return to practice