LevSMS — Full Thesis from Sam (Voice Transcript)

Date: 2026-05-13 ~5:45 PM EDT
Source: Sam voice memo dictated while reading Codex's LevSMS Router Go-Live Briefing
Status: raw thesis — to be tightened by Codex (LevSMS router owner) and Sam over next sessions
Why this lives here: voice was rich enough to count as a spec source, not just a comment

Verbatim themes (paraphrased for readability; voice intent preserved)

1. Twilio status confirmed

• Campaign was non-good → in progress → APPROVED yesterday
• Successfully sent SMS to Sam's 917 personal number — received
• WhatsApp test worked
• Inbound webhook works in principle; sheet structure + permission logic still to be built

2. Permissions / phone number authorization (the core architecture)

• HIS phone number = full access · everyone else = scoped access by tags
• Sheet aggregates: phone number → tags → permission level → allowed triggers
• Anti-spoofing required:
• Empty/blank "From" must be REJECTED, not given default access
• SMS sender ID can be spoofed — need out-of-band confirmation for sensitive triggers
• Two-factor auth for premium / age-gated content
• Whitelist by phone number, with CONTROL tab + REFERENCE_RULES tab governing
• Example: "I do not allow any of them for sports" — phone number tag "no-sports"
• Premium tiers: bronze · gold · diamond · etc. — per-tier menu items unlocked

3. Menu paging logic

• User gets 5 of 50 options based on tags/permissions
• Numbered 1-5 even though they're #8, #33, #39 in master
• Dynamic per user

4. ZIP code intelligence (the LOCATION layer)

• ZIP IN system → return weather + Hebrew date + parsha + Shabbos candles + zmanim + local headlines
• ZIP NOT in system → polite "not in my system, would you like me to add it?"
• User-add → triggers sheet update + crawl-for-that-ZIP
• Whitelist mechanism for new ZIPs (Sam approves before public)

5. Pilot scope

• North Woodmere kosher community first
• Davening / prayer times (zmanim) per ZIP
• Weather + Hebrew date + parsha
• Shabbos candle-lighting
• Local synagogue + restaurant + service info
• Sources to crawl over time: yeshivaworld.com, theshivaworld.com, local NYC orthodox community sites

6. Two-factor auth model

• Sensitive trigger → SMS confirm via reply OR different channel
• Counter-spoof: confirm via different number/email
• Age-gated: parent consent for minors
• Premium tiers may require email + phone double confirm

7. Character limits / monetization seed

• SMS = 160 chars per segment
• Reserve some for ads (future monetization)
• Smart truncation
• "Sponsored by" footer line for free tier

8. RIDESHARING (the NEW big build)

• Sam doesn't want WhatsApp chat-spam ("I'm going to X, who needs ride?")
• Doesn't want to be the de facto chauffeur of his community
• LevSMS-driven private channel:
• User texts "going to [destination] [time]"
• LevSMS keeps registry of who needs rides
• Connects them directly (peer-to-peer match)
• No public group chat noise
• Trust signal: only people Sam pre-tagged as "approved drivers" can offer rides
• Riders can be more open

9. Cadence + transcription workflow problem (the META gap Sam flagged)

• Trying to record on iPhone but uses iPad mini sometimes
• Need to pick ONE cadence: phone OR iPad OR computer mic
• Wants comments/feedback workflow on briefings → trigger more automations
• Wants Claude Code-like trigger automations (think "Claude Code triggers" pattern from Cursor)
• Morning briefing → 3-5 questions → Sam voice-responds → triggers update

10. Brokerage alert consolidation (separate but mentioned)

• Set up alert properties across Schwab, Fidelity, IBKR/thinkorswim
• Folder routing for alerts
• Reduce dependence on Google Finance for live prices
• Pull live prices from IBKR (or Fidelity) API as primary source
• $100 to keep IBKR account active for API access
• Add Alpaca as another option
• Click "Refresh" in Google Sheet → pulls live data

11. Command Center / portal naming

• Need better name than "Command Center"
• Both Sam + Mildred login spot
• Briefing link for Mildred
• Pre-given access controls
• Backend permission controls
• Hacker-proof

12. The macro/micro audit ask

• "Take a macro micro look at anything and everything from every repo and everywhere and anywhere"

Where this slots into existing work

Things Sam is NOT seeing (my critique — see briefing)

1. Cash Action Board guardrail violation (his own rule from this morning)
2. Ridesharing legal liability (commercial drivers, insurance)
3. Twilio costs scale fast (per-message billing eats margin at scale)
4. Sheet-as-database doesn't scale past ~10k rows (kosher community DB needs real DB at scale)
5. Voice transcription cadence problem has a simple answer: pick ONE tool, always use it
6. Anti-spoofing is harder than it sounds (requires OOB confirm, can't trust caller ID)

See outputs/2026-05-13_17-45_briefing_levsms-thesis-critique-and-real-moves.html for full critique.

Coordination note

This thesis is voice-only. Codex owns the LevSMS router build (per repo levsms-router-apps-script/). When Codex picks this up:
1. Read this doc end-to-end
2. Decide which themes go into router CODE vs router SHEET vs new docs
3. Surface concrete sheet-tab schemas
4. Push the schema decisions back here for cross-agent visibility

— Captured by Desktop Claude Code · session 34 · 2026-05-13 17:45 EDT