BOS / Obligations — Foundation Research

Deep audit · the whole obligations system, code- and data-verified
Generated: Tuesday, June 30 2026 · 4:53 PM EDT (NY) · Brain (🧠) of the Inbox
Sources: 3 parallel research threads — (1) architecture/code, (2) live sheet data (v1 + v2 read live), (3) build history + briefings. Cross-checked vs OBLIGATIONS.md, PROPERTY_UTILITIES_MAP, chanie-bills, memories.

You were right: BOS isn't "asleep," it's half-built and unplugged. You have the canonical data (v1), the smart planning logic (v2), and a finished consolidation engine (v3) that was built to fix a broken interest formula — and never run. You pulled the triggers on purpose so it would stop emailing wrong numbers, and the fix has sat un-executed since. So nothing "works" today not because it's broken beyond repair, but because the last step never happened. Meanwhile the live data hides three genuinely urgent money facts (below) that no system is currently watching.

1 · The system — v1 / v2 / v3 / the "full repo"

Layer	What it is	State
v1 — "Monthly Obligations" `1L_rxCSOnc…`	The canonical obligations ledger + the reporting engine (Daily/Weekly/Monthly + Ops Summary + wife report + Mildred handoff). Pulls from HS Unified + Eden OS + Daily Tracker + Quick Capture.	DORMANT — triggers off since ~5/27 (your call)
v2 — "Monthly Obligations v2" `1qNYUbbvaLxpO3…`	Where the smart logic actually lives: snowball/debt classification, an 18-month cashflow projection, seasonality, strategy scaffolding. But it's the duplicate-email source (overlapping AM/PM, no dedup).	DORMANT — flagged to retire (harvest its planning tabs first)
v3 — Obligations rebuild `v3-obligations.js` (inside v1)	The intended fix: collapses the 27-tab sprawl → ~9, installments first-class, "Snowball Plan — High APR First," 18-month runway. Migrates from v1 + a Rocket Money CSV. A finished runbook exists.	BUILT, NEVER RUN — manual menu, no run steps executed
command-center (the "full repo")	An abandoned Next.js + Supabase dashboard (auth/obligations/tasks/reports), ~40% done, never deployed. Its schema was harvested into v3's design — not lost, folded in.	ABANDONED — graveyard candidate

Why it went dark: the old Sunday email had a sign-flip bug on debt balances + a dead interest formula (off a 22,475-row Rocket Money CSV). You disabled the triggers deliberately to stop wrong-number emails ahead of the v3 cutover — and the CF-cron morning brief became the "one place" in the meantime. (Memory fix: the v2 sheet ID in project_bos.md was dead/wrong — corrected to the live id.)

2 · The money truth — three urgent facts no system is watching

🔴 Your HELOC is negative-amortizing. US Alliance, ~$217,000 @ 8.50%. Monthly interest ≈ $1,537 but the minimum is $1,248.50 — paying the minimum grows the balance ~$289/month. This is the single most important debt fact in the whole system, and it's not on any dashboard you look at.

🔴 Your two highest-rate cards are invisible. Amex Business SPG (29.99%) and Barclay AAdvantage (28.49%) have no balance loaded ("NEEDS BALANCE — paste CSV"). The avalanche can't be real while the worst exposures are blind. Loaded card debt ≈ $62.8K; true total is higher and currently unknown.

🔴 Tomorrow (Jul 1): National Grid gas $356.77 + Kia $599 are unscheduled / no pay-from. Darchei $2,281.82 + BYAM $1,100 auto-pull from Chase 7792 — needs Chanie to fund it.

And the sheet lies about safety: v1's cashflow tab claims net +$5,469/mo. Reality (cash-inflow memory): liquid cash under $100, pinches met with Fidelity margin. The "net positive" is fiction because the income input was never set. Don't trust the green number.

3 · The debts — rate + balance (best-verified)

Debt	Balance	Rate	Monthly	Pay-from
Fifth Third — home (589)	−$383,168	needs APR	$3,312.59	Chase 5007
US Alliance HELOC ⚠ neg-am	−$217,000	8.50%	$1,248.50	Chase 5007
Selene — 9312 FL	—	needs APR	$3,379.91	9312-6651
Shellpoint — 9332 FL	—	needs APR	$3,161.51	unconfirmed
SBA loan	−$40,900	0%	$113.61	TBD
Honda / Kia (auto)	−$11,400 / −$14,376	0% / 0%	$475 / $599	1260? / unset
Citi 2550 (Chanie)	−$14,710	14.99%	$528.53	—
US Bank Busn 0049	−$14,104	19.49%*	$239–365	—
Chase Ink 4089 / Freedom 8412	−$11,933 / −$7,590	16.74% / 18.49%	$288 / $194	—
Amex SPG / Barclay (no bal!)	UNKNOWN	29.99% / 28.49%	—	Chase 5609

*US Bank 0049 = a card (19.49%) carrying a fixed 4.90% installment plan; the $239 plan payment is the live obligation. Loaded credit-card debt ≈ $62.8K (excludes the blind Amex/Barclay/Simplicity).

Strategy: tagged "snowball," but the real plan is avalanche + balance transfers

Rows are tagged snowball (smallest first); v1's dedicated tab implements avalanche (highest-APR first: Amex → Barclay → Citi Simplicity → US Bank → Freedom → Ink → Citi 2550 → HELOC). v2's strategy template is scaffolded but unauthored — no single ordering is locked.
The real-world move (credit-stack memory, Jun 11): ~$15.2K of 0% balance-transfer capacity (BofA $9K @ 0% to ~Apr 2028 + Chase Slate $6,191) — aim it at the 28–30% Amex/Barclay to kill interest, not principal. And stop new card apps (each inquiry hurts the HELOC/refi that's the actual cash prize).

4 · Tuition, installments, utilities, business/personal

Tuition (Jul 1, Chase 7792, autopay): Darchei $2,281.82 · BYAM $1,100 · Arievim $341.67. Camps: Agudah $200×2/mo, Ohr Chedva $250 (not scheduled). Installments (Affirm, ~$848/mo from card 0405→5609, Apr-stale — re-verify): 6 plans, ~$4,105 balance; Klarna should be closed (May 1). Utilities: home — Liberty Water $330.64 (auto, 5609), National Grid $356.77 (🔴 unset), PSEG ~$405 (manual); FL — Duke + Toho (auto, but Toho 9312 likely mis-routed from 9332-2528). Split (v2): Business ≈ $10,423/mo (13 rows, HSC + FL STR), Personal ≈ $15,750/mo (23 rows) — encoded in the Type column + by pay-from account.

Monthly outflow: ~$25K (v1 $25,332 · v2 $25,109 · OBLIG.md $24,567) — but v1 internally conflicts ($24,531 cashflow tab vs $16,650 totals tab). Reconcile before quoting any number.

5 · Conflicts to settle (data is dirty)

National Grid $356.77 (trust v1+property map) vs $831.77 (stale v2).
Mortgage pay-from Chase 5007 (trust v1) vs 5777 (v2).
Selene/Shellpoint pay-from swapped — trust v1/portal: Selene = 9312-6651; Shellpoint unconfirmed (needs proof).
FL mortgage amounts — trust v1/portal $3,161.51 / $3,379.91 (v2 stale).
Honda pay-from — card-map (CSV-verified) says acct 1260, not 7792.
Sheets are full of #REF!/#VALUE! + epoch-garbage dates + stale cached rows. The data needs one clean pass.

Recommendation — don't revive the broken emails; finish the fix

The old v1/v2 trigger emails are the wrong thing to turn back on (broken formula, dup noise, dirty data). v3 already is the answer you built — it just never ran. The path:

Load the truth. Paste a current Rocket Money / Chase CSV so the 5 blind cards (esp. Amex 29.99% + Barclay 28.49%) get real balances. Without this the avalanche is guesswork. (5 min, yours.)

Run v3. Execute the runbook on the v1 sheet: V3_BUILD_STRUCTURE → V3_MIGRATE_FROM_LEGACY → harvest v2's cashflow/seasonality tabs → fill gaps. Collapses 27 tabs → 9, installments first-class, real runway. (I drive it; you click the menu items it needs.)

Wire it into the spine, not an email. v3's cashflow + "what's due / paid / needs you" feeds the morning brief + a home tile — a live view where you already look, cloud-reachable with your PC off. Retire v2 + the command-center repo.

Lock one strategy. Avalanche on the 28–30% cards using the ~$15.2K BT capacity; make a real plan for the negative-am HELOC (minimum is losing money); stop new card apps. Settle the pay-from conflicts so nothing slips.

Step 1 + 2 are the unlock. Everything else (the live wiring, the strategy) builds on a clean v3. Say go and I start by speccing the exact run sequence + the CSV you need.

Source trail
File: outputs/2026-06-30_16-53_research_bos-obligations-foundation.html
Built from 3 code/data-verified research passes (architecture, live v1+v2 sheets, build history) · live sheets read 2026-06-30 · ops-api ad740523 · working dir: 05. 2026 BH
Canonical IDs — v1: 1L_rxCSOncM6WtqCYaEbooNdxiXbfOi_Sfm7k0IAAsBo · v2 (corrected): 1qNYUbbvaLxpO3DfUiwmb3_3jLvrTnAYNjKeuje7qTl4 · v3 = v3-obligations.js inside v1's script project. Numbers are best-verified reads of dirty sheets — treat as the map, confirm a figure before wiring it.