MIS v2 — Next Steps (SCALE-FIRST: build for 200+, prove at 2)

Repaired 2026-05-27 per Sam: "We're only at 2 tickers and perfecting here, but think over 200 immediately — so whatever it looks like now has to look good then. Make it look good on the moon/Mars so it's good on Earth."

The principle: every surface + the data path must scale linearly from 2 → 200+ with NO redesign. The 2-ticker version is the small case of the final thing, never a throwaway. If a surface would break or get ugly at 200, it's the wrong surface — fix it now while it's cheap.

What this CHANGES about the design

Stacked cards are NOT the primary surface. Today's dashboard stacks one card per ticker — fine at 2-5, an endless scroll at 200. Retire card-stacking as the primary; it becomes the drill-down.

Four-tier surface (looks identical at 2 or 200):
1. Glance bar (fixed size — never grows): market row (S&P/Dow/Nasdaq/VIX/10Y/Gold/BTC) + sector rotation (11 bars) + a one-line portfolio status (up X%, drawdown, vs 2.5% target).
2. Shortlist — "what to do today" (fixed ~3-5 rows): the thesis gate — top names by SACS, max 2/sector, earnings-safe, capital-capped. This is how 200 scored tickers compress to an action list. The answer to "does it tell me what to do."
3. Universe table (2 rows now, 200 later — same component): one row per Include=Y ticker — verdict badge · grade · SACS · today % (color) · 30-day mini-spark · sector. Sortable + filterable (by verdict / sector / grade). Adding tickers = adding rows, nothing else.
4. Drill-down = the Test Entry (full card + chart + how-to) on tap of any row.

Data path built for scale (mostly already true): the engine computes once → FINAL_STATE_ENGINE table → every surface reads it (no per-render recompute). Finnhub on a rolling 40-stalest/run. Full SVG charts only on the shortlist + drill-down (200 inline charts = too heavy); the universe table uses a tiny trend mark, the chart appears on tap.

Repaired next-step sequence (each step notes WHY it scales)

1. Universe table view — the scale-ready surface. Build it NOW at 2 tickers; it's a 2-row table that becomes a 200-row table untouched. This is the scale insurance + the direct answer to "how's it look with more tickers." ← do FIRST.
2. Shortlist / "Today's Trades" gate — top 3-5 by SACS, max 2/sector, no earnings within 7d, capital-capped 20%, total risk ≤5% (from the requirements checklist). Scales because 200 in → 5 out is the whole point.
3. Bot ↔ Telegram (Phase 2) — ask it from your phone: dashboard / shortlist / a ticker / "what changed." Scales because the bot reads the same precomputed tables.
4. Pushed briefing + more/less dial — a brief that auto-arrives (morning), with a control for more/less detail. The "where's the briefing" answer; ties the morning-spine. Scales because it's a render of the shortlist + glance bar.
5. Thesis / cash-flow layer — monthly 2.5% tracker, drawdown gauge (−5/−10/−15/−20), A/B/C bucket report, extraction recommendation, sector pie (from MIS_V2_REQUIREMENTS_CHECKLIST.md). Portfolio-level, size-independent.
6. THEN load the ~200-ticker universe — flip the universe size into the now-proven, scale-ready surfaces. No rebuild — just more rows + the rolling refresh doing its job.
7. After scale: real performance/margin (your Fidelity Positions CSV), self-learning auto-accumulation (daily history trigger), deeper per-ticker charts.

Internalized 2026-05-28 NIGHT — additional cross-session signal (Plaid live, MCPs, emoji fix)

Caught up on the post-evening parallel-session shipments before clasp:

• Plaid is LIVE on a new ops-api Cloudflare Worker (separate from portal + command-inbox Workers). /link.html bank-connect flow live; /plaid/balances endpoint live; Tartan linked + tested; production access pending Plaid review (sandbox today). Home page Cash tile is now a LIVE balance pull (commit 34271d8). For MIS, this unlocks an optional augmentation: portfolio metrics could merge HOLDINGS_CLEAN's broker cash with /plaid/balances for a complete cash picture. Captured in plaid-ops-api-live; not wired into MIS yet (no permission given).
• 5 MCPs installed by the parallel-session: Chrome DevTools + Playwright + GitHub + Cloudflare + Twilio. Closes the "visual-test gap" — future MIS HTML can be Playwright-validated before deploy.
• Bot emoji ?? bug FIXED (commit aba21ec): sendTelegram_ now defaults to plain text (drops parse_mode=Markdown). My MIS bridge (mis-bridge.gs) responses are already plain-text-safe; the rule still favors plain ASCII for tg.ps1 multiline pushes.
• Mobile-rendering skill shipped with 13 rules; DESIGN_SYSTEM.md §6 updated today. MIS surfaces are partially compliant (font stack + body + tabular-num + safe-area + viewport-fit landed earlier today). The container queries on the universe table land today's push fully into the scale-ready bracket.
• Queue consolidated 68 → 8 projects + auto-triage spec. The MIS Gap Pulse already plays into this queue model — when it asks Sam a question, the answer can flow back into the queue board.
• 9 urgent cards surfaced from the inbox sweep (Darchei registration LATE, Tello deactivated, HOA waivers, school billing). Not MIS items but contextual: Sam has competing urgencies; MIS shouldn't ping during those.

For when clasp lights: deploy stages 1-10 as queued. Then offer the Plaid-balance augmentation + Playwright visual test as deliberate next steps (Sam's call). Don't pile them on automatically — they're scope additions, not pending work.

Internalized 2026-05-28 PM — cross-session signal (Pulse @47, DESIGN_SYSTEM, bot eyes/memory)

Caught up on the parallel-session work via .remember/now.md, today-2026-05-28.md, and the last 40 commits. What that surfaced + applied to MIS v2:

DESIGN_SYSTEM.md (locked 2026-05-27, mobile rules added 2026-05-28) is the canonical visual language for EVERY surface in the workspace — MIS, BOS, Eden, portal, briefings. Rules: forced light · dark header band + gold accent · bubbly cards · 4px spacing grid · 15px/1.4 body · tabular-num feature on number columns · 12px card padding (not 16-20) · pills ≤14 chars (long state → 8px colored dot + inline text) · tables collapse to cards on <640px via container queries · viewport-fit=cover + env(safe-area-inset-*) · system font stack (-apple-system, BlinkMacSystemFont, "SF Pro Text"). Memory: design-system-canonical.

MIS surfaces now compliant (committed this turn, deploys with clasp):
- ✅ System font stack + 15px body + 1.4 line-height + tabular-num on the outer wrapper
- ✅ viewport-fit=cover + color-scheme=light meta on HtmlService output
- ✅ env(safe-area-inset-*) padding on the outer wrapper
- 🟡 Pills longer than 14 chars in places (10-DAY · 2WK) — switch to dot+text
- 🟡 Tables not yet container-query collapse on <640px (universe table + holdings will need this for ≥30 rows on phone)
- 🟡 Card padding mix (some 14px instead of 12px) — sweep

Bot upgrade signal (bot-eyes-and-memory-ops): the command-inbox bot now has eyes (search_queue, get_card) + persistent memory (remember op + Brain_Feed cross-session log + 24-fact context tab). Pulse @47 is live in observe mode with the ask-first-on-money/legal/family rule. This is important for MIS because:
- The MIS Gap Pulse I just built shares the same tg.ps1 webhook pipe — proven cadence channel.
- The MIS bridge (command-inbox/mis-bridge.gs) is intentionally additive over the bot's new eyes — don't duplicate the memory layer; bot keeps conversational memory, MIS keeps trade-system state.
- The ask-first-on-money/legal/family rule aligns 1:1 with MIS "advisory, never autonomous" + the Wash-Sale gate (don't auto-execute, surface for Sam's call).

Mobile-rendering skill (hookstreet-skills/mobile-rendering/) exists as a callable skill — when v2 surfaces grow (universe table at 200 rows, holdings on phone), use it to enforce DESIGN_SYSTEM compliance rather than hand-rolling each render.

mis-daily-report skill also exists in hookstreet-skills/ — should be the canonical place for the daily brief composition once the cadence layer ships.

Telegram = plain ASCII (per docs/CHANNELS.md + the channels memory). Bullets • and middots · are borderline; the safer default for tg.ps1 messages going forward is - and , so it never renders as ??.

Internalized 2026-05-28 — re-reading the SmartHub origin chat (March 2025)

Deep re-read of archive/smarthub-history/2025-03-17_smarthub-origin-chat.txt. The 4 system properties Sam locked a year before MIS v2 are still the right test — and v2 is uneven on them:

Promoted into the gap analysis as a result:
- Wash-sale tracking — the JPM "can't re-enter" case from March 2025 lifted to TOP-priority (gap #0 in MIS_V2_GAP_ANALYSIS.md). New canonical reject code WASH_SALE_BLOCK. Real $ consequences.
- Pattern recognition from trading history — extends TRACK RECORD beyond "what did MIS say before" to "what setups have worked for ME before" (read TRADE_LOG by setup type + verdict, compute hit rate). Phase 3.
- Shabbos / Yom Tov calendar gate — no auto-briefs / no auto-trade-imports on chag; the morning brief still fires Saturday but says "Market closed - Shabbos." Small but explicitly part of "spiritually attuned."
- Proactive blind-spot calling — new behavioral memory proactive-blindspots-and-bigger-thinking. I should be the one to surface the blind spot before Sam asks for it (the gap analysis itself was an example).

Today's 10x build (2026-05-28 — "go build my man" round)

Built locally + committed; everything below ships the moment !clasp login runs.

✅ Engine (MIS v2 _gen.py / Code.gs):
- New tabs: Ideas, Thesis_Log, Questions, TRADE_LOG
- misV2Decide_ / misV2GetDecisions_ + tap-to-log buttons + progress bar on tickets
- misV2Idea_ / misV2Thesis_ / misV2Question_ + auto ticker + theme extraction (extractTicker_, extractTheme_)
- misV2Upload_ generic upload endpoint
- misV2ImportTrades_ — Fidelity + Schwab execution-email parser, dedup by Order_Id (you don't type a fill)
- misV2GapPulse_ — scans for open decisions near target/stop, held names without thesis, unanswered Qs, FSE ADDs with no decision; returns questions for Sam
- misV2GapPulseAndNotify_ — pushes those questions to Telegram via the notify webhook
- misV2DailyBrief_ — text brief (market · shortlist · open decisions · gaps); detail = less/normal/more
- misV2BriefAndNotify_ — pushes the brief to Telegram on a trigger
- misV2Shortlist_ — the "what to do today" gate (top 5 by SACS, max 2/sector, earnings-safe)
- misV2UniverseHtml_ — universe TABLE (scale-ready 2→200, one row per ticker, verdict-sorted then SACS)
- misV2ShortlistHtml_ — shortlist as a sheet table
- Dashboard now renders: SHORTLIST → tickers → sectors → UNIVERSE TABLE
- Daily triggers: setupTradesTrigger (5:30 PM trade import) · setupGapPulseTrigger (7:45 AM pulse) · setupBriefTriggers (7 AM + 4:15 PM briefs) · setupAllTriggers (one-shot)
- Endpoints: fn=idea | thesis | question | upload | trades | pulse | gaps | brief | briefpush | briefdetail | triggers | shortlist | universe
- Manifest updated: gmail.readonly + script.scriptapp + script.send_mail (one reauth covers all)

✅ Bot (command-inbox/mis-bridge.gs + 4 cases in start-here.gs):
- 9 MIS commands: MIS / MIS <TKR> / MIS HOLDINGS / MIS SHORTLIST / MIS DASHBOARD / MIS BRIEF [MORE|LESS|NORMAL] / MIS PULSE / MIS GAPS / MIS TRADES
- 3 capture commands: IDEA: <text> / THESIS: <text> / Q: <text> → write to MIS sheet via HTTP
- HTTP-based bridge (no cross-script sheet access) — additive, existing handlers untouched

📋 EXACT DEPLOY RITUAL (when you reauth — ~3 minutes total)

1. !clasp login — 30 sec, browser opens, sign in with Google.
2. I run (~2 min): clasp push --force + clasp deploy on both MIS/v2 and command-inbox. I'll notify when each lands.
3. First time you tap the MIS link — Google asks to grant the new scopes (Gmail read + triggers + mail). Tap accept. ~20 sec.
4. You set 2 Script Properties on the command-inbox script (so the bot can call MIS): MIS_V2_URL + MIS_V2_TOKEN. I'll give exact values.
5. You tap a link I send: …&fn=triggers — creates all 5 daily triggers (Finnhub refresh, history, trade import, gap pulse, briefs). 5 sec.
6. You text the bot MIS in Telegram. It replies with today's market + shortlist + link. Bridge confirmed.

After that: morning + EOD briefs auto-arrive · gap pulse fires ~7:45 AM · trades auto-import after close · you can IDEA: / THESIS: / Q: the bot anytime · the test entry is actionable · the universe table is scale-ready.

Today's additions (2026-05-28 — "moving" round)

• Actionable Test Entry — built, awaiting deploy. A Decisions tab + tap-to-log buttons (TOOK / PASSED / TRIM / CLOSED WIN / CLOSED LOSS) on every ticket + a progress bar for open positions (% to target · pull to stop). Decisions persist in the sheet → the next time the ticket opens, the buttons show history + status. Code committed; deploy blocked on clasp login (auth rolled over).
• Cloudflare memory: not needed. The sheet IS the persistent decision/state layer (free, you already own, auto-syncs). Use Cloudflare for the portal's delivery + Access gate (already in place); use the sheet for memory. Don't double-store.
• Synced design language across surfaces (sheet rendering, portal, dashboard, entry, future bot replies): one palette + one chip/badge/bar vocabulary. Today: light card, gold accent (#d4a72c), green/red direction badges, verdict colors locked (ADD #15803d · STARTER #1e40af · WATCH #b45309 · REJECTED #b91c1c · EARNINGS_RISK #7c3aed). Lock these as the MIS design tokens + apply consistently in the universe table + future surfaces.
• Progress bars become a first-class element (already used in 52w range + new decision progress) — also apply to: monthly 2.5% target, drawdown gauge, bucket allocation vs target, sector concentration. One visual idiom; readable everywhere.
• Where's the briefing? It's the dashboard today (always-there glance). A pushed brief with a more/less dial is queued under "pushed briefing + control" (step 4).

The discipline that doesn't change

• Perfect the brain + each surface at 2 tickers (INTC/NVDA) before loading 200. Scaling = changing the universe size, not rebuilding the surfaces.
• Advisory, never autonomous · verdict depends on intent/horizon · no false numbers (feedback_mis_advisory_not_autonomous).
• Glanceable, self-explaining, charts-not-prose (the design language locked today).

Where things stand right now (so the plan starts from truth)

✅ Brain (FSE) at 2 tickers · Test Entry (upside, tweakable, track record) · Dashboard (market + ticker charts + sector rotation) · real gold/BTC · holdings consolidated (5 accounts) · live bookmarkable link (fn=dashboard&view=1).
⏳ Universe table · shortlist gate · bot wiring · pushed briefing + controls · thesis/bucket layer · performance/margin · then scale.

Living doc — update as steps ship. Companion: MIS_V2_STATUS.md (where we are), MIS_V2_REQUIREMENTS_CHECKLIST.md (everything asked), MIS_V2_TRADE_INGESTION_AND_BEHAVIORAL.md (holdings/behavioral capture).