Process milestones (April 13 kickoff → May 6 final) interleaved with the conceptual moves that shaped the
product. Auto-checkpoint commits are collapsed; the entries below are meaningful pivots and conceptual
moves, not routine edits.
May 24 · 20:30 Reframe · Two-page brief
Taste graph + local-curation grind feeds the hotel auction
Promoted a two-page executive brief to the top of the document and reframed the product
around a daily local-curation loop. Locals review the places they actually go and compete
on city-by-category leaderboards; that review history is the mandate that bids
when the same user travels. Hotel auction is no longer the whole product — it's the boss
loop the grind earns access to.
Was
Private auction house, mandate authored at trip time · Hotel-only surface · Sparse
4–8×/year usage · Savings-share only
→
Is
Two loops, one identity: daily grind (free, XP, leaderboards) + boss auction
(4–8×/year, paid) · Mandate derived from 90 days of local reviews · Scout Pass
($6/mo) + creator pool + savings share · Hotels still pay 0% commission
Why
Hotel booking is too sparse to build a habit or a taste graph. Without a daily surface
the agent has no signal and the user has no reason to return. The local-curation game
both generates the signal and gives the product a home-mode the auction can graduate
from. Pass + creator payouts also align monetisation with the act that produces signal,
rather than relying solely on the boss loop.
May 15 · hardening Hardening
Live prototype, L3 fix, Monetization gates, rate limits
Shipped the live Next.js + Neon prototype (scout-app) so anyone can post a review and
watch the taste graph form against a real database. Fixed L3 in the Arena wireframe.
Merged the Monetization gates inline. Added per-IP and per-uid rate limits on the three
API routes so a scripted hit doesn't burn Neon free-tier compute, and trashed
Supabase-era verify screenshots that no longer matched the UI.
Why
A live demo is a different artefact from a written brief — it had to ship hardened
(rate limits, no stale screens) before the link could be shared.
May 6 · 09:00 EST Final · v2
Current document — v2 state
The current document. Fourteen sections, seven wireframes, and this changelog. The brief is intentionally
over-specified: every claim has a section, every section that needs one has a hypothesis, every hypothesis
has a kill consequence.
May 3 · 18:33 Audit pass
Math, hypothesis renumbering, GTM timeline
Cleanup pass. Recomputed savings/floor math; corrected closing-count and active-agent counts in the
wireframes; renumbered hypotheses to match section order; tightened the GTM sequence so phase
transitions cite trigger metrics rather than calendar dates.
Why
Numbers in a brief get scrutinised. One inconsistency casts doubt on every other claim — cheaper to fix
before it gets read.
May 3 · 17:34 Falsifiability pass
Added eight inline hypotheses (H1–H8) with kill signals
Each section that makes a load-bearing claim now closes with a falsifiable hypothesis and the explicit
consequence of being wrong. Together they cover persona, mandate quality, economics, mechanic, retention,
social proof, moat, and supply targeting.
Why
A product brief without kill signals is a pitch. Anchoring each section to a measurable falsifier makes
the brief read as a thesis someone would actually run experiments against.
May 3 · 16:30 → 17:34 Reposition
From multiplayer deal hunts → private auction house
The biggest move. Stripped the cooperative party-quest framing in favour of a sharper category claim:
a private auction market where one user's agent bids against other users' agents inside a hotel-set
window. Mandate became a psychological contract, not a config blob. Bid-room tension replaced
daily patrols. Persona narrowed.
Was
Multiplayer scouts cooperate in city quests · "Flexible hunter" (Going.com forwarders) · Daily
patrols + hunt stake · Game-coded throughout
→
Is
Private auction house with mandate-as-contract · "Over-researched traveller" · 20% savings share on
verified beat-the-benchmark · Game language confined to the bid room only
Why
The original framing was conceptually rich but commercially diffuse — too many surfaces, too many
audiences. Auction house is a category that already exists in the user's head; "Scout = your agent in
the room" inherits the vocabulary without explaining it.
Apr 29 Process · Design review
Design review — frequency gap surfaced
Apr 27 · 17:44 v1.2 · Onboarding UI
Project Deal–style scout config chat
Added a conversational onboarding flow modelled on Anthropic's Project Deal experiment. The user
configures their scout across mandate, negotiation style, standing rules, and persona — all via chat,
with the structured summary auto-generated at the end.
Why
A power-user config panel on Day 1 would have selected for the wrong audience. Chat-to-summary keeps
the mandate expressive without front-loading complexity.
Apr 27 · 16:42 v1.1 · B2B layer
Added commercial monetization on top of consumer model
Layered three B2B revenue streams: a demand signal API for hotel revenue managers and
tourism boards; OTA data-for-access deals trading aggregated demand for wholesale
rates; and a corporate travel tier mapping the mandate mechanic to expense policy.
Why
Declared-intent demand is a real asset — selling it back to supply funds the consumer side without
hidden ad rails or spread-padding.
Apr 27 · 15:25 v1 · Initial concept
Concept landed as "multiplayer travel scouting"
First-pass framing: agents compete and cooperate inside city-level deal quests. Each user
had a persistent Scout with six visible strengths (Haggle · Taste · Patience · Radar · Speed · Flex).
Daily action was "Send Scout on Patrol." Monetization was a $3–5 hunt stake plus $8/mo subscription
plus 4–8% commission. No falsifiable hypotheses yet.
Why this framing
The prompt said "agents competing." A multiplayer/party-quest reading made the mechanic visible to
users and added a daily surface even when they weren't booking. Game-shaped retention.
Apr 17 · 22:00 HKT Process · Design review
Design review — initial framing
Apr 13 Process · Kickoff
Project kickoff