Board v2 handoff (10 Oct 2026)

What changed since the founding session (9 Oct), why, and where things stand. Read after 2026-10-09-founding-session-handoff.md.

Stevan's 10 Oct brief (dictated)

He rejected v1 of the board: titles didn't resonate ("Neutral vendor evaluation pack... I don't get it"), descriptions assumed context, first steps hid meta-work ("pick the top five corridors. I don't know what they are"), some items weren't projects at all (senior routing), and 19 flat items had no focus. He asked for: plain language, a visible deliverable per item ("show the end result"), small concrete steps, wireframe thumbnails to react to, cards with click-to-open instead of a long scroll, research into how great partner programs actually run (his words: "run a few agents"), stress-testing by subagents, and concentration on THREE bets:

1. Spain property estate agents with Tom Culshaw, CallumAI as the free-tool wedge ("a high IQ, 4D chess move").
2. Dubai corporate service providers, hook = the new-company bank-account pain, he can broker introductions.
3. Media agencies via Marcus's ambassador network (Wow: Marcus, Imran, Nathan, David Morley, Peter Czapp).

What was built

• Board v2: three dark "bet" tiles up top (click to filter), card grid with wireframe SVG mocks per deliverable, click-to-open modal with What it is / What you end up with / The steps / Why it matters, rating buttons on card and modal, notes in modal. 21 items, 5 lanes (spain 3, dubai 2, media 2, foundations 11, relationship 3), 17 to rate, suggested-now concentrated to 7: SP-2, DX-2, MA-2, F-5, F-9, F-11, R-1.
• Four Opus research agents ran; digests saved verbatim in docs/research/ (FX partner programmes, programs outside payments, Dubai CSP channel, Spain estate-agent channel). The wiki on hermes supplied the media-vertical context (concepts/ct-media-agency-vertical.md; Marcus runs client money through Wise and hates it; Imran's override model; Nico's affiliate book is internal proof).
• An Opus stress-test agent then reviewed the full draft against the research. Its major catches, all applied:
  • Dubai legal framing: the UAE trigger is promotion (licensed activity since Sept 2026), not just commission; DX-2 now asks "may anyone acting for CT promote to UAE companies at all", freezes all outreach until the memo (due 31 Oct), sequences spend (providers free first, counsel only on a provider yes), and pre-builds the no-go branch (UK formation agents; no FX specialist in their partner directories).
  • Sell-first-verify-later: media pitch and offer merged into MA-2 (prove on Marcus's agency, then arm ambassadors); SP-2 signing gated on SP-1's legal pack; F-9's published rate gated on an economics model (Lumon's 0.276%-of-volume benchmark would eat CT margin on large flows; Wow overrides add a second paid layer).
  • Factual fixes: Francois has STEVAN's number, not vice versa (Stevan has him on LinkedIn only): also corrected in global memory; F-5 reframed around the held-payment reality plus a same-day hold-explanation rule; R-3 updated to introducer-first (white-label later); a dozen overclaims reworded (e.g. "CSPs bring us every client they form", "receive client money in any currency", "every partner reaches a first client inside a month").
  • Added F-11 (provider capability matrix) replacing five scattered "Paul confirms" steps; DX-1 merged into DX-3.
  • Declined with reasons: did not cut F-7 (Stevan's own idea; demoted later with the counterargument in src_detail), did not merge R-2 into R-1 (long game stays visible), did not merge F-2/F-3 (distinct owners/deliverables).
• Kill-or-scale numbers now live in the bets: Spain = FX opt-ins per agency at pilot day 60 (needs 3-4 FX buyers/agency/yr to pay); Dubai = gate memo by 31 Oct; media = funded agency accounts from the first five intros by quarter end.

State at save

Deployed to https://ct-partnerable-lk0q-om10-j6jo.pages.dev (live sync verified, rating round-trip tested, KV reset to empty). Stevan has NOT rated v2. Nothing sent to Francois yet. SIMED Málaga (12-14 Nov) and GITEX (7-11 Dec) are the near calendar hooks; Peter/Wow meeting 19 Oct.

Resume

1. node scripts/sync-ratings.mjs for ratings; his rating beats suggested.
2. The research digests are the evidence base; cite them rather than re-researching.
3. Item copy renders on the page: plain sentences, no em-dashes, keep the honesty bar the stress test set (no claim that doesn't trace to the provider matrix or a digest).

Post-rating corrections (10 Oct, evening)

• Stevan rated 10 items (5 now, 4 next, SP-3 parked). He offered 7 back for refinement instead of parking them.
• F-5 deleted: his correction: Francois's payment issue was unrelated to regular partner payouts, which run fine. The 9 Oct inference (payout reliability as program gap) was wrong.
• F-4 deleted after a code check: the CT app already enforces corridor-aware fields. bank_account_required_fields.yml drives per-country/currency bank fields; the allocate-funds component asks the backend per recipient whether purpose of payment is required (getPaymentPurposeData → is_required + corridor options) and validates it; purpose_code flows to Currencycloud (tcc_v2 mapper). Residual question (do dealer-booked trades get the same check, across all 7 provider integrations?) folded into F-2's gap step.
• R-2 parked at his instruction (focus dilution), with a revisit-after-Francois-win note in KV.
• F-2, F-3, F-9, R-1 refined per his notes: F-2 reframed as a gap-check for US corporates (he says onboarding already covers most of this); F-3 reframed as portal country guides with data-picked countries; F-9 narrowed to "ammunition for fiduciary channels" with a kill test; R-1 carries the verdict he asked for (not a long shot on the win, slow on timing; Paul pre-reviews). All four gained a "How others already do it" modal section with the concrete research examples; these are the only 4 left to rate.