MindSet
Roadmap

Where this is going.

Reset 2026-07-15 off the Detached Coach vision. MindSet's real product is a structured mirror that becomes a detached second-self coach. Six short-term loops close the gaps that stand between today's Signal Bus and that coach. Five long-term phases build on top of it.

  1. 01Shipped

    Accounts + every Set synced, with editable Outcome Models

    • Email and Google sign-in.
    • All nine Sets sync across devices — Daily Record, MindSet Outcome Model (MOM), Weekly Review, Values, Take Stock, Reorient, Trigger Repair Set (TRS), Anchors, Gratitude.
    • MOM sizes each outcome (quick win → transformation) and drafts a weekly roadmap with belief, state, and daily action cues.
    • MOM roadmaps are fully editable — add, edit, reorder, or delete milestones, weekly focus, and daily actions. Lock when ready to feed today's Daily Record.
    • Cross-account safety on shared browsers — unstamped local data is wiped before any cloud sync.
  2. 02Shipped

    Phase 2 polish

    • Shipped 2026-10-02 — the Second-Self Set can now answer "what might happen if…" questions. It finds the times you were in a similar spot before and shows what followed in the days after, using your ratings, felt senses and evening writing. It never predicts with certainty and never tells you what to do.
    • Shipped 2026-09-25 — Asking the Second-Self Set a question now searches everything you've written for the people and words in it, and shows which entries, by date, the answer drew on.
    • Shipped 2026-09-25 — Luminous Coherence sessions now carry your written words into the Signal Bus, not just the fact that a session happened. Your insights and notes from past sessions were read in too, and the Second-Self Set knows Luminous Coherence is a meditation practice.
    • Shipped 2026-09-25 — a new Set, MindSet Second-Self (MSS), at /sets/second-self. One page to ask what your own record shows: a quick tip, today, this week, this month, last month against this one, all time, or a question you type yourself. Every answer is drawn from your entries and says how many it read and which days it covered. Nothing you ask is saved.
    • Shipped 2026-09-25 — your evening writing now feeds the Signal Bus. What went well, what was hard, what you'd do differently, gratitude and your practice-lens answer all become signals, and so do the lens answers in Luminous Coherence sessions. Past entries were read in too. The "MindSet sees you" card names a lens you keep working, or what your evenings keep calling hard, only once it has shown up at least twice.
    • Shipped 2026-09-25 — Luminous Coherence sessions show a panel for each practice you have an active model built on: its posture in one line and a question you can answer, saved with the session. The session overview shows the same postures as a quiet reminder.
    • Shipped 2026-09-25 — the Daily Record carries your practice lens. Suggestions from an active model built on a practice show a small lens tag, and the evening reflection adds one question from that practice beside gratitude. With no practice-based model active, nothing changes.
    • Shipped 2026-09-15 — Luminous Coherence session tiles now show a clear gold fill, border and checkmark when you select them. Before, tapping a center, quality, wave rating or mind mode barely changed its appearance — the color was so faint it was hard to tell anything happened. Fixed at the root so the same gold tint works everywhere it's used.
    • Shipped 2026-09-15 — the Daily Record page was tidied behind the scenes: the same six sections, prompts and look-back panel, now built from smaller named pieces instead of one very long file, so future changes are safer. Nothing was removed and nothing moved on screen. Also checked live: asking Swing Thoughts for a suggestion and keeping one writes the cue to your account and records it on the Signal Bus.
    • Shipped 2026-09-10 — Take Stock now shows the right day. Your entry was always saved on the correct date, but the date printed on the page was reading a day early. Fixed everywhere Take Stock shows a date, including your history list.
    • Shipped 2026-09-08 — a clean-up pass, plus two real fixes it uncovered. Your Daily Oath now lives on your account instead of on one device: the wording you write and the choice to show it on your Daily Record follow you to your phone, and the oath no longer quietly saves a second copy of itself into Swing Thoughts when you open MindSet somewhere new. The oath strip also appears and disappears the moment you flip the switch, instead of waiting for a reload. Behind the scenes, the Set and signal counts on the Overview and this page are now counted from the real registries, so they can't drift out of date, and Relationships is marked live.
    • Shipped 2026-09-08 — The Life I Want, a new Set at /sets/life-i-want. Your four-tier manifesto (Inner Peace → Purpose & Value Creation → Community → Mastery) is now worked, not just read: each hill is a real Outcome Model with its own beliefs, actions, practice lens and AI roadmap, built bottom-up on purpose — you can't start Hill 4 before the ground under it exists. Hills carry an honest Working / Unworked state, and their locked roadmaps feed the Daily Record like any other Model. The Daily Oath lives at the top of the page, editable, with an optional toggle that renders it as a quiet strip on your Daily Record — and it auto-seeds itself into your Swing Thoughts so it's always within reach. A new oath_shown signal (once a day, only when the strip actually renders) lets the bus eventually answer whether oath days go better. Bus now 37 kinds / 15 producers / 17 Sets.
    • All Set pages refresh from the cloud on every visit (2026-06-13).
    • Security hardening (2026-06-13): AI features locked to signed-in users; cloud-sync validators tightened.
    • Running Lists sync across devices (2026-06-10).
    • Second-Self mirror strip now reads on Simple Equation and Weekly Review (2026-07-26), in addition to Daily Record. Same silent-first behavior — if there's nothing honest to say, or if you've turned MindSet's use of the mirror off, the strip stays hidden.
    • Daily Record (2026-06-10): AM and PM Top 3 above Suggestions; new Personal + Work Running To-Do Lists.
    • Cloud saves retry quietly + show a single short note on final failure (2026-06-10).
    • MindSet Simple Equation — configurable cards, settings synced across devices (2026-06-01).
    • Simple Equation reality-first layout (2026-07-16): weight sliders hidden behind Advanced; default view shows live scores and drill-downs only; 'What is this?' explainer distinguishes real data from configurable weights.
    • Stoicism added as a Known Practice (2026-07-21): new page at /practices/stoicism with the Dichotomy of Control, Four Cardinal Virtues, Amor Fati, and Premeditatio Malorum. Values Compass gets a one-tap 'Load the Four Virtues' button. An opt-in Stoic overlays toggle surfaces a reflective sub-prompt inside Reorient, Trigger Repair Step 2, Weekly Review Lessons, and Daily Record Section F — additive only, nothing changes if you leave it off. Email verification and leaked-password protection turned on the same day.
    • Buddhism added as a Known Practice (2026-07-25): new page at /practices/buddhism with the Four Noble Truths and the Eightfold Path. Opt-in Buddhist overlays add reflective sub-prompts inside Trigger Repair Steps 2 and 3 (naming suffering, seeing craving/aversion), Reorient (grasping vs. letting go), Weekly Review Lessons (clinging vs. release), Daily Record Section F (Right Mindfulness), and Simple Equation Friction drill (Right Effort). Additive only.
    • Taoism added as a Known Practice (2026-07-25): new page at /practices/taoism with the Three Treasures and wu wei. Opt-in Taoist overlays add reflective sub-prompts inside Reorient (act or yield), Trigger Repair Step 4 (least-force response), Weekly Review Lessons (effortless vs. grinding), Daily Record Section F (wu wei flow-check), and Simple Equation Attitude tile (fragile grinding vs. aligned effort). Additive only.
    • MindSet Simple Equation — honest formula states (2026-07-21): the Personal Formula tile now surfaces negative-lift drags (behaviors that co-occur with lower peak-day rates), soft-pedals the sentence when fewer than 21 days are logged, calls out ties when the top two drivers are neck-and-neck, and handles the all-consistent case where nothing has a real split. Client-only tightening — no new AI, no new data.
    • Shipped 2026-09-03 — MindSet can now be connected to an AI assistant (Claude, ChatGPT, Cursor) as a set of tools. After you sign in and approve the connection on a consent screen, the assistant can read your recent Daily Records, your Signal Bus signals, your to-dos and your swing thoughts, and add a to-do or keep a new swing thought — always acting as you, never as anyone else. Nothing is reachable without your approval, and you can deny or revoke it.
    • Shipped 2026-08-30 — the Daily Record flows in the order you actually use it. The day's entry comes first: Anchors, Top 3, Reflection, Ratings, Primer, straight under the header, with Felt sense and Sleep now labelled sub-blocks inside Ratings. A thin bar sticks to the top of the page showing A to F chips that fill in as you write, plus your save state and Export — tap a letter to jump there. Everything MindSet wants to flag you — the coach line, reorient prompts, repairs in flight, the mirror, emotional weather — is grouped into one collapsed "things to look at" card instead of a stack, and it disappears entirely when there is nothing to say. Everything backward-looking — streak, the 30-day grid, anchor hit-rate, active Models, your To-Do summary and recent days — moved into a single "Your record" panel below the entry, open on desktop and collapsed on mobile. Nothing was removed.
    • Shipped 2026-08-30 — if you signed in but never finished setup, the Daily Record now tells you. A gold Finish setup strip sits at the top of the page with a one-tap path back into the three-step welcome, and a Not now if you'd rather get on with your day — it only asks once. The moment you finish setup, anywhere, it disappears.
    • Shipped 2026-08-30 — first-run is visual now, and it can't be skipped by accident. If you signed up, confirmed your email and then came back through the sign-in page, MindSet used to drop you straight onto the Daily Record with no setup and no instructions — that is fixed: anyone who hasn't finished setup lands in the three-step welcome instead. The welcome itself reads as pictures rather than paragraphs: a progress rail showing Anchors, Values and Land, short lines instead of blocks of text, example value words, and a closing board showing what you just set up. And on the Daily Record, new members see a Start here strip — four numbered steps that jump you to Anchors, Top 3, Felt sense and Ratings — which fades away once you've logged five days. There's a Redo setup link in Account if you skipped it.
    • Shipped 2026-08-12 — Luminous Coherence is now a practice you can actually do inside MindSet. It joins the other five Known Practices at /practices/luminous-coherence — the two realities, the three centers, the four stages, and what coherence feels like when it is real. It also has its own Set: a forty-session arc at /sets/luminous, ten sessions per stage, with a short daily logger for what you noticed, what loosened, and the one small choice that came out of it. There is no streak to break — miss a day and the arc simply waits. Each stage closes with a reflection that opens once you have logged all ten. You can switch on a Luminous overlay across every Set the way you can with Stoicism, Buddhism or Taoism, and you can build an Outcome Model on it. Luminous Coherence™ is Rod Stryker’s work (© 2025) and appears here with his permission.
    • Shipped 2026-08-12 — all five Known Practices can now carry an Outcome Model, not just two. Mind · Body · Spirit and The Six Principles were the only practices a MindSet Outcome Model (MOM) could be built on; Stoicism, Buddhism and Taoism could only add a reflective line on top. Every Model now has a “Built on” choice, and the practice you pick is the lens the AI roadmap is written through — a Stoic Model separates what is yours to control from what is not, a Buddhist Model works with what is being clung to, a Taoist Model looks for the move that takes least force. Same four quadrants, different lens. Each practice page also has a “Start a Model on this practice” link, and existing Models keep reading as The Six Principles unless you change them. Live-verified 2026-08-30 — switching a Model's lens and regenerating its roadmap produces language that matches the chosen practice.
    • Shipped 2026-08-11 — the partner overview is current again. The /overview briefing (and the PDF leave-behind) was still describing fifteen Sets and eleven signal producers, from before Swing Thoughts and To-Do Lists shipped. It now lists sixteen Sets with their practices and overlays, names twelve producing Sets on the signal-bus diagram, and states 31 signal kinds. Fresh PDF: MindSet_Overview_Alpha_Partner_v3.pdf.
    • Shipped 2026-08-11 — To-Do Lists is its own Set at /sets/to-do. The Personal and Work running lists moved out of the Daily Record and onto their own page, with more room to work. Two changes you'll feel: checking something off no longer erases it — completed items stay behind a 'Show completed' line with the date you finished them, per list — and you can put an item back if you tick it by mistake. The Daily Record now shows a single line telling you how many items are carrying, with a link across.
    • Shipped 2026-08-11 — Swing Thoughts, a new Set at /sets/swing-thoughts. A place to keep the cues you actually want within reach. Three lists: short one-liners you can glance at mid-day, longer passages for the thinking behind them, and AI suggestions drawn from what MindSet already knows about you — your recent signals, your felt senses, your Outcome Models and your Values. Nothing is saved until you keep it, and if there isn't enough history yet it says so rather than handing you a platitude. Pin, edit, archive, delete. No daily ask.
    • Shipped 2026-08-11 — a tidy-up pass under the hood. Nothing you see changed, but three things got safer. The code-quality checks went from 1,637 complaints to zero errors, so a real problem can actually be seen now instead of hiding in the noise. The rule that decides which felt sense represents your day now lives in exactly one place — two copies of that rule is how the lost entries happened on 7 August. And the internal maintenance jobs (re-reading your history in the current vocabulary, rebuilding what MindSet knows about you) now sit behind one door with a plain description of what each one does and when to run it.
    • Shipped 2026-08-11 — two things that were quietly lying to you, fixed. Felt senses logged after 7 August weren't being saved: the entry sat in the page waiting on a delayed save, and a sync from your account replaced it first. Logging now saves the moment you add it, and the card says "Saved ✓" so you can see it landed. Second, the Attitude Index was hammering you with a follow-through gap you hadn't earned — completed actions weren't matched back to the intent that created them, gaps only refreshed overnight, and five open items maxed the penalty no matter how much you'd done. Now a completed action clears its own intent by identity, gaps refresh when you open the Simple Equation, up to 20% slippage costs nothing, and every open item has Done / Drop buttons. Same days, honest number: 71 Steady.
    • Shipped 2026-08-03 — Felt sense now asks why, not just what and where. Logging a felt sense adds what set it off (a thought, a memory, a person, a situation, something physical, or out of nowhere) plus an optional line for the thought or memory itself. That line is read the same way a journal entry is, so a belief or a person named in it lands on the bus with the felt sense. What MindSet knows about you and the mirror strip both gain a "what set it off" read — quiet until three triggers are logged.
    • Shipped 2026-08-03 — Felt sense replaces the old body check-in on the Daily Record. The question used to be "where did today's strongest feeling sit in your body?" — one tap, once a day, and no name for what the feeling actually was. Now you log a felt sense: what it was (joy, calm, energized, warmth, pride · stress, anxiety, frustration, sadness, heaviness, tightness, pain, numbness), where it sat, and how loud it was. One a day is the ask; log as many as you notice. Each one is its own signal on the bus, carrying whether it was carrying you or costing you — so MindSet can finally say "stress in your shoulders, 6 of the last 14 days" instead of just "your chest". Slept well and Deep-work block are unchanged. The mirror strip and What MindSet knows about you both read the new pattern, and stay quiet until there are at least three logged.
    • Shipped 2026-08-03 — your history now reads in the richer vocabulary. When signals gained the "need" and "stance" dimensions, only new entries carried them; everything you had already written stayed in the old shape. A one-shot replay rebuilt 895 signals from every Set you've used — anchors, journal, Reorient, Trigger Repair, Outcome Models, Take Stock — so Insights, the coach and the Simple Equation now read your whole history the way they read today. Nothing you wrote changed; MindSet just re-read it.
    • Bugfix: Signal Bus vocabulary cleanup (2026-08-02). The /overview diagram was listing three signal kinds that don't exist on the bus. It now shows only real kinds plus an honest '29 kinds, a sample below · +19 more' caption, and a regression test fails if a fake kind ever slips back in.
    • Trigger Repair (2026-06-01): Step 6 carry-forward + Patterns view.
    • MindSet Simple Equation MasterSet (SET-014, 2026-06-01).
    • Weekly Review W7 · AI synthesis (2026-06-01).
    • Reorient mood chips + Quiet reflection panel (2026-06-01).
    • Take Stock Then-vs-Now % change + carry-forward stitch (2026-06-01).
    • Bugfix: Weekly Review surfaces past weeks automatically (2026-06-01).
    • Bugfix: Daily Record completion + streak count any meaningful entry (2026-06-01).
  3. 03Now

    Sets that talk — the Signal Bus

    • Shipped 2026-06-22 — Loop 1: a shared Signal Bus that every Set now writes to (Take Stock, TRS, Reorient, Gratitude, Weekly Review, MOM). One canonical vocabulary of life areas, beliefs, emotions and body locations means your Sets stop being silos.
    • Shipped 2026-06-22 — Loop 2: the Insight Engine is now visible. Weekly Review has a new W8 · Cross-Set Insights card (7-day window). Trigger Repair and Reorient each have a 60-day Pattern panel scoped to that Set. The MindSet Simple Equation has a Friction by life area panel (30 days). Gratitude has a Gratitude by area panel.
    • Shipped 2026-07-02 — Signal Bus health pass: first live verification against your account's data. Fixed two bugs (Weekly Review autosave was creating dozens of duplicate signals per save; journal mood free-text was polluting the emotion list). Cleaned up historical duplicates. Bus is now 1-signal-per-entry with a clean canonical emotion set.
    • Shipped 2026-07-03 — Loop 3.1: TRS carry-forward 'Anchor' now writes a real todo into today's Morning Spirit anchor in Daily Record (with dedupe + cross-device sync). And a new Daily Record Section G — a journal area with two spaces: General Journal and BE MY OWN HERO — both feed the Journal Set and the Signal Bus, tagged daily_record so cross-Set insights pick them up.
    • Shipped 2026-07-03 — Loop 3.2: TRS carry-forward for MOM belief and Values check-in now performs real writes. Picking 'MOM belief' appends the belief to your active Outcome Model. Picking 'Values check-in' logs a check-in on your first value. If the target Set is empty when you pick it, nothing is lost — it's saved on the repair and shows up as a small 'From Trigger Repair' banner on that Set's page with one-click Attach or Dismiss.
    • Shipped 2026-07-05 — Loop 3.3: Repair Status board at /sets/repair-status. One page for every open handoff from a Trigger Repair (MOM beliefs + Values check-ins) with Attach or Dismiss, a Recurring Beliefs view (any belief that has come back 2+ times in the last 30 days, with a 'Take to MOM' shortcut), and a Recently Attached list. Linked from the Trigger Repair page header.
    • Shipped 2026-07-05 — Loop 3.5: Reorient nudge on the Daily Record. When a belief or life area has come up 3+ times in the last 7 days across your Sets, a soft mint panel appears above Anchors: 'This has come up a lot — a quick Reorient?' with one click to /sets/reorient or Not today. First surface where the Signal Bus reads back to you in the moment, instead of just collecting.
    • Shipped 2026-07-05 — Loop 3.6-measurement: The Reorient nudge now records both when it's shown and when it's followed (two new signal kinds). A small tile on the Reorient page mirrors it back: 'You've been nudged to Reorient N times, and followed through on M.' No judgement — just the pattern, so next week's suggested experiments can be shaped by real data.
    • Shipped 2026-07-05 — Loop 3.6-full: Weekly Review now has a W9 · Suggested experiments card. When a belief or life area has come up 3+ times in the last 7 days across your Sets, you get up to 3 gentle experiments to commit to. One click and it lands in this week's review AND appends to your active MOM's Actions quadrant for next week — a real stitch from reflection into action, not just another note.
    • Shipped 2026-07-05 — Loop 3.4: Take Stock now has a Then-vs-Now view of your Signal Bus. When you take stock, Section C shows what your attention looked like in the 30 days after your Origin Snapshot versus the last 30 days — top beliefs, top life areas, top emotions, net polarity — with what's new highlighted in mint and what's dropped off struck through. Loop 3 is now complete.
  4. 04Then

    Weekly Insights digest + MOM-learns-you

    • Shipped 2026-07-05 — 4.1 Weekly Insights digest (in-app). New page at /insights/weekly gives you a Sunday-morning read of your week: top 3 patterns across all your Sets, one experiment worth trying, and one belief worth watching — with one-click shortcuts to Reorient or take it to your MOM. Linked from your Weekly Review.
    • Shipped 2026-07-05 — 4.2 MOM-learns-you. When you generate or regenerate a MOM roadmap, it now reads your last 14 days of Signal Bus (top beliefs, top life areas, top emotions) and grounds the draft in what you're actually noticing day-to-day — not just what you typed into the form. A mint '✨ Insights informed this' chip on the roadmap header shows when this happened, and hovering it lists what was used.
    • Shipped 2026-07-26 — 4.2 Thread C · MOM-learns-you deepened. The same Second-Self read you see in the mirror strip is now used when MindSet drafts a MOM roadmap — and it says so. The chip is now '✨ Grounded in your last 14 days'; hovering it lists your dominant emotions, any beliefs the journal keeps flagging but you haven't named yet, your peak-day pattern, and where feeling lands in your body. The prompt also gained a soft rule: if your recent pattern is heavy or a belief keeps surfacing, Week 1 leans toward re-anchoring and naming, not raw ambition. Honors your 'Let MindSet use this when it talks to me' toggle.
    • Shipped 2026-07-31 — Coach nudges tell the truth about what you've already done. The 'coach noticed' card reads your weekly Second-Self snapshot, which meant a belief you named in a MOM mid-week kept getting flagged as unnamed for up to seven more days — muting it only silenced it until midnight. Now the card re-checks your actual MOM belief tags at the moment it loads, so anything already named never fires. And 'Name it →' now genuinely names it: clicking adds the belief to your most recent MOM instead of just walking you to Simple Equation. Observational nudges (loudest emotion, peak-day pattern) still dismiss for the day — they're notices, not to-dos.
    • Shipped 2026-08-01 — Shadow Work (new Set). MindSet already spots the beliefs your journal keeps flagging that appear in no Outcome Model and no Value — beliefs you live but have never claimed. Until now it only pointed at them. Shadow Work takes one at a time through four steps: say it in your own words, name what the belief has given you (the gift comes first, on purpose), count what it costs you now, then own it in a sentence and pick one small experiment for the week. Finishing names the belief on your most recent Outcome Model and drops the experiment into this week's Weekly Review, so it lands in the follow-through loop you already keep. Reachable from the Identity Coherence card on Simple Equation, or directly at /sets/shadow-work.
    • Shipped 2026-08-01 — Shadow Work now opens for any loud belief, not only unclaimed ones. On day one the Set only offered beliefs your journal flags that no Outcome Model names — but naming a belief takes one tap, so the page went empty before a single session was ever run. Corrected: every belief your journal has flagged at least twice in the last 30 days is workable, loudest first, in two lists — 'Unnamed and loud' and 'Named but never worked'. Naming a belief is the doorway; working it is the room. Finishing an already-named belief still writes the session and drops the experiment into this week's review. And if a belief you've worked keeps showing up, you can work it again with your earlier writing kept in front of you.
    • Shipped 2026-08-01 — Shadow Work experiments now surface inside Weekly Review with a Done / Missed resolution. The experiment you picked in Shadow Work appears in its own card, and clicking Done or Missed writes a signal the intent-vs-behaviour loop reads as a resolution. That closes the loop from 'I named a belief' to 'I did something about it this week'.
    • Shipped 2026-08-01 — Values Compass now writes to the Signal Bus. Naming or renaming a value records it once, and a new 'Lived this today' button on each value logs a same-day check-in, so alignment is something you mark rather than something inferred. Ten Sets now emit; Relationships, Simple Equation and the Anchors history stay read-only views over the same stream by design.
    • Shipped 2026-07-10 — Track A round 1: added Signal Bus emitters to Anchors and Running Lists, plus AI extraction that reads every Journal entry and pulls beliefs / life areas / emotion clusters constrained to MindSet's canonical vocabulary. Silent on failure, invisible in the UI.
    • Shipped 2026-07-10 — Track A revised: after 9 days of your daily use showed 0 new signals had actually persisted, moved emission from the browser (unreliable fire-and-forget) to the server (atomic with each Set's own save). Every Set — Anchors, Running Lists, Reorient, Trigger Repair, Take Stock, MOM, Journal — now writes its signal in the same authenticated call as the Set row, so nothing gets lost to page navigations or auth races. Also added a hidden one-shot backfill so your 9 days of history land in the bus immediately.
    • Cut 2026-07-12 — Realtime cross-device sync. You use one device at a time; the token is better spent producing meaning-signal than syncing between phantom devices. Revisit when there's a second user.
    • Deferred to 4.4 — real Sunday-morning email delivery of the digest. The in-app version at /insights/weekly is where you'll read it for now.
  5. 05Next

    Meaning over mechanism

    • Approved direction 2026-07-12 after a critical read of the Signal Bus. 331 signals over 30 days, but 81% are checkbox telemetry (anchor hits + MOM action completions). Only 6 belief-tagged rows in the whole bus. Every downstream reader — the Reorient nudge, the Weekly Digest, MOM-learns-you — was built assuming rich belief/emotion data on a stream that's mostly noise. This phase rebalances toward meaning.
    • Shipped 2026-07-12 — 5.2 Belief extraction fixed + backfilled. Prior to today, the AI was pulling only 1 belief from 36 journal entries because the prompt was over-conservative and never saw the belief tags in context. Rewrote it to lead with inference (not literal-match), gave each belief tag a concrete anchor phrase so the model has something to reason against, and now feed it your explicit belief/mood fields when you write them. Added a lexical fallback so obvious patterns aren't lost. Then ran the new extractor over your existing journal (hidden dev route): 29 of 31 entries came back with a belief — 94% yield, 52 beliefs total, 83 signal rows written. The belief layer of the Signal Bus is real for the first time.
    • Shipped 2026-07-12 — 5.3 Emotion Weather. First surface that reflects emotion back to you. A mint sparkline of the last 14 days of journal emotion signals shows up above Daily Record's Anchors section and inside your Weekly Review (W8b). It calls the dominant emotion (sadness, calm, love, pride, anger, shame, joy, fear, overwhelm, or numbness), and a small chip says whether that emotion is up, down, or steady versus the prior week. No judgement — a weather report. Silent when there's nothing to show. Reads what your journal is already producing (post-Slice C) — no new writes, no schema change.
    • 5.1 queued — Split the bus into two layers: behavior (anchor hits, action completions) vs. meaning (beliefs, emotions, triggers, reorients, journal insights). Every insight surface defaults to meaning; Weekly Digest gets a separate small Consistency tile for behavior.
    • 5.4 queued — Recalibrate the Reorient nudge. Current 3×/7d belief threshold has fired exactly once in 30 days. Either lower it or trigger on emotion clusters instead.
  6. 06Now

    SET-018 Relationships

    • Family + Relationships is the largest tagged area in your Signal Bus (22 signals) and no Set owns it. Trigger Repair and Reorient are the nearest surfaces, but relational patterns don't have a home.
    • Shipped 2026-07-12 — Slice B: Relationships extraction. Every journal entry now runs through a second reader (mirrors the belief extractor) that pulls the people you mentioned and how each one felt (supportive, strained, neutral, ambivalent, distant, or conflict). Each person becomes one relationship_mention row on the Signal Bus, first-name-only, deduped per entry, idempotent so re-runs don't double-write. A hidden dev route reprocesses the 31 existing entries. Same wiring and logging conventions as Slice A/C — no schema change, no new tables.
    • Shipped 2026-07-12 — SET-018 Relationships log surface at /sets/relationships. Read-only view of the people showing up in your journal: dominant felt quality per person, mention count, last-mention age, secondary-quality chips, plus a Recent Mentions feed. Configurable window (30 / 90 / 180 days). Methodology anchor: Buddhist metta (loving-kindness) — noticing who's present and extending goodwill. Chosen over Stoic oikeiosis because the felt-quality categories already map cleanly to metta's four brahmavihārā.
    • Shipped 2026-07-13 — Simple Equation MasterSet reconfigured to consume the Signal Bus (SET-014 v2). Framework now weighs belief clarity and relationship health alongside anchors, MOM, values, and Weekly Review. Consistency adds a system-usage cadence score (days you touched any Set beyond the Daily Record). Friction adds emotional load (share of heavy emotion clusters) and ignored reorient nudges. New default 'signal-weighted' formula adds up to +5 each for supportive relationships and follow-through on committed experiments. The AI 'one move' now targets the softest sub-driver, grounded in your top belief and dominant heavy emotion. Signed-out users still get the pre-bus computation.
    • Shipped 2026-07-16 — Manual person merge on /sets/relationships. Duplicates like Sam/Samson or PSG/Patricia can now be collapsed into one canonical person via a per-card menu. Aliases are stored per-user in a new person_alias table with row-level security, applied at read time so nothing is written to your existing journal signals — unmerge is fully reversible. Merged cards show a '+N merged' pill listing the folded aliases.
    • Queued next — Weekly Digest relationships tile (dominant person + felt-quality shift week-over-week), and a Reorient relational branch that surfaces a relational prompt when recent mentions skew strained / conflict / distant.
  7. 07Absorbed

    Apathy detector — folded into the Short-Term Track

    • Reset 2026-07-15. The apathy detector is no longer a separate phase. It's what LOOPS S1 (intent-gap), S2 (meaning-density tile), and S3 (embodied check-in) collectively become when they stitch together.
  8. SNext

    Short-Term Track — Detached Coach foundations

    • Reset 2026-07-15 off the vision Detached Coach. MindSet's real product is a structured mirror that becomes a detached second-self coach — one that sees you more clearly than you see yourself because it's not inside your moment. Six gaps stand between today's bus-fed hub and that coach. Six loops close them.
    • LOOP-S1 · Intent-vs-Behavior signal. ✅ Shipped 2026-07-15 (MVP). A nightly server job compares what you said you'd do (locked MOM daily actions, Daily Record tomorrow primers, Weekly Review experiments — aged 3–21 days) with the behavior signals that followed. Non-matches become intent_gap rows on the Signal Bus; matches on later runs clear the stale gap. The Simple Equation Friction panel shows follow-through gap count and the most recent stated intent as evidence. S1.1 ✅ Shipped 2026-07-15 — the Reorient nudge on your Daily Record now has two variants: if 2 or more intents you named recently are still open, it renders 'you said [intent], then didn't — want to unpack why?' and outranks the tag-spike variant. Closes the intent-vs-behavior gap.
    • LOOP-S2 · Meaning-density tile on Weekly Digest. ✅ Shipped 2026-07-15. New tile on /insights/weekly scores how meaning-loaded your journal entries have been in the last 14 days (emotions + beliefs the extractor found, per entry) with a this-half-vs-prior shift chip. Silent when you haven't journaled yet. No new writes. Closes the meaning-density gap.
    • LOOP-S3 · Embodied check-in on Daily Record Section E. ✅ Shipped 2026-07-15. One-tap body-location chip (chest / stomach / throat / head / shoulders / whole body) beside the existing ratings, plus a light / moderate / strong intensity row when a location is picked. Emits an embodied_state signal per day so the body vocabulary now populates every day, not only during Trigger Repair. Closes the embodied-state gap.
    • LOOP-S4 · Identity coherence score. ✅ Shipped 2026-07-15. New Identity Coherence tile on the MindSet Simple Equation compares three belief surfaces (Values name+description, MOM Beliefs quadrant, journal-extracted belief_add signals over 30 days). Returns three chip groups: 'named and lived' (overlap — the honest part), 'journal keeps flagging, unnamed' (the loud finding — beliefs to consider adding to your MOM), and 'named but quiet'. Score = overlap ÷ lived, 0-100. Silent when you have no MOM, no Values, and no journal beliefs yet. No new tables, no new signal kinds. Closes the identity-coherence gap.
    • LOOP-S4.1 · AI-suggest belief tags on MOM save. ✅ Shipped 2026-07-15. Closes the honest gap S4 surfaced: the coherence tile read 0/100 because MOMs are usually written aspirationally ('I am capable') and the regex only fires on limiting phrasing ('not enough'). Now when you edit your MOM Beliefs quadrant, a small AI call reads the prose and proposes 0–3 canonical belief tags (from the 8: not-enough, must-be-perfect, invisible, no-control, not-safe, not-lovable, can't-trust, responsible-for-others). A single confirm bar appears — 'AI heard: X · Y' with Accept / Edit / Dismiss — and accepted tags become what Identity Coherence reads. Empty is a valid answer (AI doesn't force tags). One AI call per meaningful Beliefs edit; hash-guarded so tweaking other quadrants doesn't re-fire. Values Compass intentionally not touched (values are aspirational, beliefs are limiting — different reflective mode). Manual '+ Tag beliefs' link still available when AI returns nothing.
    • LOOP-S4.2 · One-tap Claim from the Identity Coherence tile. ✅ Shipped 2026-07-21. Each chip in the 'Journal keeps flagging, unnamed' column now has a small ＋ button — tap it and that belief is named on your most-recent MOM, the chip moves to 'Named and lived', and the score updates on the spot. A 5-second Undo toast is right there if you tap by mistake. If you don't have a MOM yet, the toast points you to /sets/mom. Naming is not healing — it's Path A. Path B (the belief stops firing in the journal) is the real work; this affordance just makes Path A one tap instead of a detour.
    • MSE-V2 · Personal peak-day formula. ✅ Shipped 2026-07-21. /sets/simple-equation now leads with a derived one-sentence formula for your peak days — learned from your own Daily Records. MindSet looks at eight things you do each day (3+ anchors hit, AM Top-3 filled, PM Top-3 filled, reflection written, embodied check-in, tomorrow primer, anchor notes, gratitude), compares peak days (overall ≥ 4, or your top-tercile days if peaks are rare) against non-peak days, and picks the three behaviors most associated with your peaks — with an optional 'gate' if one behavior stands out. Renders as: 'X + Y + Z, gated by W.' The Framework × Consistency − Friction section is still there below as 'Practice Health' — a diagnostic of the system, not the formula. Needs 14+ logged days and 3+ peak days before it renders. This is not a causal proof — it's a first-honest cut; real regression, situation-specific formulas (project/transition/goal), sleep + deep-block first-class inputs, formula-drift alerts, and AI narration are all on the roadmap.
    • LOOP-S5 · Temporal signature. ✅ Shipped 2026-07-15. New heatmap tile on /insights/weekly shows day-of-week × time-of-day (morning / afternoon / evening / night) with your peak reflection window called out. Read: your natural rhythm — not a scoreboard. Closes the temporal-signature gap.
    • LOOP-S6 · Second-self memory v0. ✅ Shipped 2026-07-21. MindSet now keeps a compact weekly snapshot of who you are right now — beliefs you've named vs beliefs the journal keeps flagging, dominant emotions, your peak-day pattern, where feeling most often lands in the body, open follow-through gaps, and this week's activity. That snapshot is silently prepended to every AI surface (Weekly Synthesis, MOM roadmap regen, MSE 'one move', Reorient reframe, journal signal extraction, belief-tag proposer) so the coach sees the person, not just the request. Rebuilt automatically every Sunday at 04:00 UTC; a hidden /dev/build-second-self page can rebuild yours on demand. Silent-on-empty — until you have a snapshot, every AI call behaves exactly as before. Completes the Short-Term-loop track (S1–S6) and closes the cumulative-user-model gap.
    • LOOP-S6.1 · 'What MindSet knows about you' surfaced. ✅ Shipped 2026-07-21. The Second-Self snapshot is now visible at /account/second-self — every section rendered in plain English, a Rebuild-now button, and a 'let MindSet use this when it talks to me' toggle. When you switch it off, every AI surface answers from the request alone. No new data, no new tables — just visibility and consent on top of S6.
    • LOOP-DC0 · Detached Coach v0 — proactive weekly nudge. ✅ Shipped 2026-07-21. First real payoff from the Second-Self. On /sets/daily-record, at the top of the page, MindSet now speaks once a week — one line, chosen deterministically from your latest snapshot. Priority: (1) 3+ open intent gaps → 'want to unpack one in Reorient?', (2) a belief the journal keeps flagging that you haven't named in a MOM → 'name it in Simple Equation', (3) a loud emotion (sadness, anxiety, anger, fear) with a body location when we have one → mirror only, (4) your peak-day pattern → mirror only, (5) quiet week → 'one anchor or one line keeps the mirror useful'. Otherwise silent — never bland filler. Dismiss hides it until next Monday. Respects the /account/second-self 'let MindSet use this in AI' toggle: switch it off and the coach goes quiet too. No new tables, no new AI calls, no cron — rule-based v0. Bridges to Phase 10 Coach Layer.
    • TRS-DEPTH-PASS · Trigger Repair depth pass. ✅ Shipped 2026-07-21. Five targeted upgrades to /sets/trigger-repair, not a rebuild. (1) Beliefs you name here now feed the Second-Self the same way journal-detected beliefs do — the 'lived' column in Identity Coherence finally sees your repair work. (2) TRS emotion chips like Rejected, Abandoned, Controlled, Trapped, Invisible, Lonely are no longer silently dropped — they now count toward your dominant-emotion picture. (3) A new 'repair written' signal fires when you actually write the repair (≥ 40 chars), so 'named the belief' and 'wrote the repair' are two different data points. (4) When your Stoic overlay is on, an optional Dichotomy of Control panel sits above Step 5's textarea — two boxes, 'Mine to hold' / 'Not mine', both skippable, both silent when the overlay is off. (5) A week later, on /sets/daily-record, MindSet asks about a repair you carried — three chips Held / Wobbled / Broke close the loop. No new AI calls, no new tables, no cron. Trust contract: nothing new speaks unless the specific conditions hold.
    • TRS-CARRY-SURFACE · Repairs in flight tile. ✅ Shipped 2026-07-23. The 7-day carry loop was one-carry-at-a-time — anything older than 8 days aged out silently. Now, on /sets/daily-record, a single 'Repairs in flight' tile shows every repair you carried that's ≥ 6 days old and still open, with Held / Wobbled / Broke on each row. Below the rows, a small footer summarizes how many handoffs are waiting to attach and how many beliefs have come back twice or more in the last 30 days, with a link to the full Repair Status board. When there's nothing to close and nothing on the board, the tile stays silent.
    • MSE-ATTITUDE-REBALANCE · Practice Health math rebalanced. ✅ Shipped 2026-07-23. The Attitude Index was scoring healthy practices as 'Low' because the old formula multiplied Framework by Consistency (so 47% consistency lopped almost half off Framework before friction was even subtracted). Three fixes: Framework and Consistency now add together with weights (0.55·Framework + 0.35·Consistency) instead of multiplying; the default friction penalty is softer (0.35 instead of 0.5); and Consistency's days-recorded score is now scoped to the window you're viewing instead of your entire lifetime — old gaps stop dragging your current streak. A new 'Why this number' line under the score shows the exact arithmetic so nothing is hidden. Same inputs, same sliders, honest aggregation.
    • REORIENT-NUDGE-INTENT-TEXT · Shipped 2026-07-25. The nudges on /sets/daily-record used to say '16 intents you said you'd do, but didn't' — no way to tell what any of them were. Now the top 3 intents themselves show up on the card, each with a 'Reorient this →' link that prefills the Reorient page, a 'Did it' button that closes the loop, and a 'Drop' button that lets it go. When you close one, the count drops on the spot — no waiting for the nightly recompute. Same for the Detached Coach weekly nudge when it's on the intent-gap rule.
    • BELIEF-SPIKE-CLEARUP · Shipped 2026-07-25. The 'this belief has come up N times in 7 days' nudge was counting one Trigger Repair entry as 6 separate hits, because a single TRS event writes several signal rows behind the scenes. Belief and tag counts are now deduplicated by day, so 'said once on Tuesday' reads as 1, not 6. A new 'Reoriented — clear this' button dismisses a spike until the belief actually returns — and running through the Reorient flow with the belief prefilled clears it automatically. Same-day hotfix: a rendering bug that showed 'This page didn't load' on /sets/daily-record was fixed too (a React hook-ordering mistake in the nudge component).
    • MSE-LIFETIME-DISPLAY-FIX · Shipped 2026-07-25. In Lifetime window mode the Consistency drill on /sets/simple-equation was showing '51/36500' for days recorded and 'System-usage cadence 51/36500 days · 0/100' — the 36500 was a placeholder for 'no upper bound', not a real denominator. Both rows now compare against the number of days since your first record, and the Signal Bus cadence sub-score is recomputed the same way, so Attitude in Lifetime mode nudges up slightly because cadence is no longer forced to 0. Windowed views (30/60/90) unchanged.
    • MSE-SLEEP-DEEPBLOCK · Shipped 2026-07-25. Sleep and deep-work-block are now first-class inputs to your peak-day formula, not just roadmap ideas. Section E of the Daily Record has two new one-tap chip pairs — 'Slept well last night?' and 'Deep-work block done today?' — both optional, both silent when un-tapped. As soon as you have 14 logged days with either mark, MindSet includes them in the lift analysis alongside anchors, tasks, reflection, embodied check-in, tomorrow primer, anchor notes, and gratitude. This is the first honest cut — hours slept and block length are deferred to a later refinement.
    • DC0-DAILY-WHISPER · Shipped 2026-07-25. The Detached Coach used to speak once a week on /sets/daily-record. Now it speaks daily — but only when a rule actually fires. The picker is still silent-first: 3+ open intent gaps, a belief the journal keeps flagging you haven't named, a loud dominant emotion, a peak-day pattern, or a quiet-week nudge. Otherwise the card doesn't render at all. Dismiss now hides the whisper for the rest of TODAY (not the whole week), so tomorrow's honest read isn't blocked by yesterday's tap.
    • SET-019-LIVE-VERIFY · Verified 2026-07-25. Merge duplicate people (Sam / Samson, Patrick / PSG) confirmed working end-to-end against your real data — Patrick shows +2 merged, Doc / You / Samson show +1 merged each, aliases resolving at read-time as designed.
    • SECOND-SELF-MIRROR-STRIP · Shipped 2026-07-25. The weekly Second-Self snapshot MindSet has been building in the background finally speaks on the page you open most. A quiet mirror card at the top of the Daily Record shows up to four honest lines: your recent cadence (anchor hits, journal entries, where your body tends to land), your current meaning shape (dominant emotion, beliefs you've named, beliefs still unclaimed), your peak-day pattern when it has one, and — in italics — the loudest belief the journal keeps flagging that you haven't named yet. Missing data drops a line silently; if there's nothing honest to say the whole card disappears. A small 'Full read →' link opens the account page. Silenced by your existing 'Use in AI' toggle, so nothing new to turn off. No new tables, no new AI calls — just showing you what MindSet was already learning.
    • FORMULA-TODAY-TILE · Shipped 2026-07-25, merged into the coach whisper 2026-07-31 (see DC0.2). Your peak-day formula from /sets/simple-equation now speaks daily. It reads your derived drivers (rolling 30 days) and shows today's status — 'you've done X/Y of what your peak days correlate with', with each driver checked or pending, and the gate driver called out when one exists. Silent until the formula has actually derived (needs 14+ days and 3+ peaks), so no speculation. Correlates-with, doesn't-cause — same honesty framing as the source tile. No new tables, no new AI calls; reuses the existing lift analysis.
    • MOM-LEARNS-YOU-THREAD-C · Verified live 2026-07-31. Confirmed against Michael's own regenerated MOM: the roadmap header shows '✨ Grounded in your last 14 days', and the generated 12-week overview genuinely cites what MindSet learned — the 'well-worn stories of invisibility' pattern, recent signals of sadness and fear alongside a desire for calm, and a Week 1 weighted toward stabilising the inner hero through observation and small relational experiments instead of raw ambition. The loop is closed: what you live feeds what MindSet drafts for you.
    • DC0.2 · One formula-aware whisper — Shipped 2026-07-31. The Daily Record had two cards talking past each other: the coach (who you are, from your 14-day mirror) and the formula tile (what works for you, from your 30-day peak-day analysis). A coach doesn't say two unrelated things. They're now one card with one voice: a single headline, one supporting line, and today's driver checklist underneath. Something you can actually close today — an unnamed belief, open intents, a quiet week — always takes the headline, with the formula becoming today's concrete lever ('One lever left today: a deep work block'). When the coach only has an observation to offer, like a loud emotion, the formula leads instead ('Your peaks hinge on one thing you haven't done yet') and the observation sits underneath as context rather than being dropped. If only one half has something honest to say, it speaks alone. If neither does, the card doesn't render at all. Dismiss still hides it for the rest of today.
    • SNAPSHOT-REFRESH-ON-MOM-SAVE · Shipped 2026-07-31. What MindSet knows about you used to be rebuilt once a week (Sunday 04:00 UTC) or whenever you pressed 'Rebuild now'. That meant the coach, the mirror strip and every AI surface could be reading a version of you up to seven days old. Now an Outcome Model change refreshes it: when your edits settle, when you accept proposed beliefs, and when you generate a roadmap. It's throttled to at most one rebuild every ten minutes so typing never triggers a pile of work, it happens quietly in the background, and if it fails nothing breaks — the weekly rebuild is still the floor. The account page now says 'refreshes when you change a model, plus Sundays 04:00 UTC'.
    • ONBOARD-1-2-3 · Onboarding, three builds. ✅ Shipped 2026-08-02. (1) First-run setup: signing up now lands on /welcome — name your five anchors, choose the values you're actually working, then open today's Daily Record. Progress is stored on your account, so a new device never asks again, and every step is skippable. (2) Guided first record: the first time you open the Daily Record, one quiet caption at a time explains sections A through F inline — nothing overlays the page, nothing blocks typing, 'Skip the tour' ends it for good. (3) Welcome modal rewritten in brand voice with a fourth step on what a Set actually is, and its dismissal now lives on your account instead of the browser.
    • SIGNAL-VOCAB-PASS · Richer signals. ✅ Shipped 2026-08-02. Every Set already emitted what you felt and where you felt it. Signals now carry two more dimensions: the need underneath it (safety, connection, autonomy, competence, rest, meaning, recognition, clarity) and the stance you took with it (approached, avoided, ruminated, repaired, numbed). The belief and body-location readers were widened too, so ordinary phrasing — 'braced', 'falling short', 'jaw clenched' — now registers instead of passing through silently. Nothing you type changes; MindSet just reads more of it. Insights gains two honest chip rows, 'What you needed' and 'What you did with it', and the whole point is the sentence the coach can finally say: you tend to go quiet when you feel unseen. Namespaced under the hood (need: / stance:) so existing life-area patterns stay exactly as they were, and a missing read stays missing rather than becoming a wrong guess. 12 new tests lock the vocabulary in place.
    • None of the six is pre-approved. Each is its own future authorization, per the 'ask on big' rule.
  9. LLater

    Long-Term Track — Phases 10–15

    • Phase 10 · Coach Layer. A persistent AI coach at /coach that reads the whole second-self model (built in S6) and responds in your own voice and methodology. Weekly opening: 'here's what I noticed this week that you probably didn't.' Voice pass (was Phase 6) is the delivery medium here — not a separate phase. Voice decision 2026-07-15: your coach learns your own idiom from your signals, journal, and edits — not a neutral methodology voice, not somebody else's voice imposed on you. The Fellowship Mirror (Phase 11) is the one surface that stays neutral, because it's cross-user by definition.
    • Phase 11 · Fellowship Mirror. Anonymized aggregate patterns across the fellowship, released only when there are 30+ real users. Format: '62% of the fellowship worked the belief not-enough this week — the ones who moved the needle did X.' Never social, never leaderboard. Uses the same closed vocabulary; no personal data in aggregates.
    • Phase 12 · Practitioner Bridge. Coaches, therapists, spiritual directors, men's-group leaders can hold your MindSet as a shared surface with your explicit consent (never the platform's default). First plausible paid tier: paid seats for practitioners, free for their clients. Requires the admin/role table that was in the old Phase 6.
    • Phase 13 · Practice Publishing. Set-authoring layer — methodology → prompt → data model → surface. Teachers publish their own Sets on the bus. The 'Substack for known practices' outcome and the only path that fulfills the Mission at scale, because it stops depending on how much one person can build alone.
    • Phase 14 · Ambient Signals. Google Calendar, Gmail, Aura/Oura, Apple Health become bus sources through the same closed vocabulary. MCP is the plumbing (expose MindSet as tools so external assistants can read and write on your behalf; inbound MCP connectors from Notion / Linear as Set sources). PWA + mobile is the delivery. All three land here — after the second-self can actually use them, not before.
    • Phase 15 · Personal Simple Equation™. A one-sentence personal formula for your peak days, derived from your own signals plus opt-in cross-user patterns. Situation-specific equations for a project, a transition, a goal.
Try Daily Record today