🧠 Decision Log
Why we did what we did. So future-us doesn’t have to reconstruct from scratch.
| Date | Decision | Context | Rationale | Alternatives |
|---|---|---|---|---|
| Sep 2026 | SIPP holding is VALL, not VWRP | Tracker and this log both recorded VWRP (0.22%). Actual holding is the Vanguard FTSE Global All-Cap UCITS ETF — LSE ticker VALL | 0.07% OCF against VWRP’s 0.22% — a third of the cost for the same global-equity job (FTSE Global All Cap vs FTSE All-World). Joe confirmed 15 Sep 2026; supersedes the Aug 2026 VWRP entry below. | Stay in VWRP |
| Sep 2026 | Cash buffer retired and deleted outright | The £10k buffer target had drifted into three disagreeing figures (£9,525 / £9,464 / £10,000) and two competing destinations for monthly surplus | Joe: “delete the whole cash buffer situation, we don’t need it any more.” Surplus flows to GIA, full stop. Net worth reports actual net cash (£9,674.90, stoozing netted) rather than a round target. | Keep a tracked buffer target |
| Aug 2026 | ISA moved to 90/10 MXWS/EMIM | Was 75/25 ACWI/JPLG. GIA (VWRP) will follow to 90/10 at the right point in time | 90% Invesco MSCI World (0.05%) + 10% iShares Core EM IMI (0.18%) — cheaper than ACWI/JPLG, same global coverage, auto-rebalance Pie keeps the split | Keep 75/25 ACWI/JPLG |
| Aug 2026 | All 21 cron jobs now zero-token no_agent scripts | 3 LLM crons violated the AGENTS.md convention; Google One verify was broken-by-design (reminder never created the task row) | Scripts are free, never drift, and can read the tracker directly. ANI check became a real calculator (reads YTD figures, computes the top-up). Google One reminder now adds a row the verifier can check. | Keep LLM for judgement-heavy one-shots |
| Aug 2026 | Cash model corrected: net liquid £9,464 across Tembo/Monmouthshire/RBS | Joe confirmed Snoop’s £7,520 was a stale aggregator view (net of stoozing, included RBS); Starling £21,750 was outdated | Tracked metric is now net liquid after stoozing. Weekly snapshot no longer subtracts stoozing debt twice — net worth re-baselined from £125.6k to ~£140k. Snoop retired as the cash metric. | Keep Snoop as aggregator |
| Aug 2026 | USER.md stripped of OpenClaw-era machinery | Stale timezone warning (Asia/Shanghai), dead framework instructions, old cron names were being injected as personal context | Misleading injected context is worse than none; the Hermes profile holds the real facts. Old file backed up to trashed/. | Leave as-is |
| Aug 2026 | Both SIPPs switched to VWRP | Nitro allocation (JPLG/FLXE/USSC) was overcomplicated | One ETF, no rebalancing, 0.22% fee. Same fund on both platforms eliminates allocation drift. | Keep Nitro with periodic rebalancing |
| Aug 2026 | T212 SIPP opened | Needed a second SIPP for flexibility | £2,500 in VWRP. T212 offers auto-invest. Combined SIPP total: £33.5k. | Single Freetrade SIPP only |
| Jul 2026 | Scottish Widows is RAS, not salary sacrifice | Tax code target was 2232L based on flawed arithmetic | SW contributions reduce ANI via tax code relief, not pre-tax salary reduction. Same mechanism as SIPP. Total gross target: £10k, not the earlier miscalculation. | — |
| Jul 2026 | SIPP increased to £420/mo net | Need £10k total gross pension (£60k→£50k ANI). Previous £350/mo wouldn’t hit target. | £420 × 8 months = £3,360 net + £840 relief = £4,200 gross SIPP. Combined with SW £2,400 = £6,600 remaining = on track for £10k. | Keep £350/mo and do a larger March top-up |
| Jul 2026 | LISA deferred to lump sum Apr 2027 | Monthly £333 wasn’t happening consistently. SIPP priority higher. | £4k lump sum before 5 Apr captures full £1k bonus. SIPP tax relief more immediate value (£525/mo pot growth vs £416 LISA). | Monthly £333 to DRS LISA |
| Jul 2026 | Freetrade for SIPP (not Trading 212) | SIPP transfer from Dodl completed | Free tier confirmed. Manual rebalancing acceptable for the 3-fund Nitro allocation. T212 doesn’t offer SIPP. | T212 (no SIPP), Dodl (higher fees) |
| Jul 2026 | Trading 212 for ISA | £83k in ISA, maxed for tax year | Auto-rebalance Pie feature (75/25 ACWI/JPLG at the time, now 90/10 MXWS/EMIM). Card cashback auto-invests. | Vanguard (higher fees), Freetrade (no auto-rebalance) |
| Jul 2026 | CGT harvest scheduled 25 March | Annual calendar event | Before 5 Apr deadline, gives time to settle. ≤£3k gain, repurchase distinct tracker. | Manual, no cron |
| Jul 2026 | ANI threshold check scheduled 1 March | Annual calendar event | Enough of the year elapsed to calculate actual figures. Room for March top-up before year end. | No check (risk of drifting into 40%) |
| Jul 2026 | Daily-tick Monday kickoff with priority scoring | Mon format was too verbose | Added domain one-liners + scored top 3 priorities + cross-tracker suggestion. Concise, actionable. | Full backlog dump every day |
| Jul 2026 | Cross-tracker intelligence auto-generated | Manual snapshot kept going stale | Weekly-housekeeping regenerates from live data every Sunday. Single source of truth. | Manual edits to cross-tracker-intelligence.md |
| Jul 2026 | ✅ Done items deleted from trackers, not marked | Joe: “If things are done don’t show them on my list!” | Completed rows removed entirely. No ✅ markers left visible. Clean lists. | Mark ✅ and leave visible |
| Jul 2026 | Fixed-period renewals → renewals-tracker, not general-todo | Subscriptions with end dates were scattered across files | Single tracker for all shopping-window reminders, handled by renewal-watchdog cron. | Keep duplicates in both trackers |
Add new decisions at the top. No need to capture every minor choice — just things future-us would want to know.
- 2026-11-04 — Contradictions settled. 46 of 48 answered after a full sweep (two
pending: insurance reminder display, Interpath reference). Money figures recalculated or
labelled rather than guessed; mortgage end fixed at 31 Aug 2028 with a February review;
legal completion 31 Jul 2026; Google One’s nudge/end/verify split into three exact dates;
dentist recorded as an annual June review; the RBS→Tembo rule retired. Full table in
.hermes/plans/2026-11-04-decisions.md.