Today we argue each arc and make the calls together. Every meeting feeds back through its transcript, and this deck becomes the run book. By Friday it is a graded record of the week.
Wed Aug 5 · 2–5pm · Spring Place 508 · Polymarket call 5pm · Mediaocean Thu 10:30 at 120 Broadway · iHeart Fri 1:30 · keys 1–9 or arrows to move · P for presenter notes · sealed v2 · re‑seal before 5pm Wed
Le Mans, 1966. Ken Miles holds the revs while everything in the car says go.
We played this at MadHive when the DSP first found traction. The moment is the same.
Le Mans, 1966. Ken Miles holds the revs down the straight while everything in the car says go, and he waits until the moment is exactly right before he drops the gear and buries the pedal. We played this clip at MadHive, right when the DSP first found traction, and it gets played again today because the moment is the same.
We have been short‑shifting on purpose for months: the research filed, the files built, the questions numbered, the rooms mapped. This week the straightaway opens. Polymarket at five today. Mediaocean at 10:30 Thursday, at their office. iHeart at 1:30 Friday.
We win where the legacy estate is the business: the very large, old‑world media enterprises the new‑media crowd ignores. Where MadHive made its money.
The wires were the newest thing on earth once. What we call old world was the boldest build of its day, and it gets the leapfrog: straight to AI, like a grid that goes 5G without ever laying copper.
Renovated this way, the old world does not catch up to digital. It passes it, and the most politically robust exchange in media keeps the budgets and builds the network.
In: questions, working product shown live, first proofs sized under the approval threshold.
Up: instrument → governed capability → the Boombox substrate, activating legacy software across a network downstream.
Win: value compounds across the network. We activate the stack they already own.
Where we play. Precise wins in the sandbox the new‑media crowd ignores: very large, old‑world‑adjacent media enterprises whose legacy estates are the business. iHeart and Mediaocean are the archetypes: vendor‑receptive cultures, enormous flow, and a hand‑made quarter‑end somewhere in the building. This is where MadHive made its money: enterprise, broadcast‑adjacent, relationship‑driven, patient once in. Polymarket is the deliberate exception, a fast‑data native that proves the machinery quickly.
How we get in. Questions are the interface: every arc runs on a numbered question table, and every answer moves a wall, opens a door, or converts a verbal yes into a date with an owner. We show the working thing and bend it live in the meeting, then let them design the yes; that is how the iHeart proof of concept got granted out loud in thirty‑two minutes. The first proof is always small and sized under the approval threshold: a retrospective on a closed campaign, shadow mode, no re‑tagging, recommendations not writes, governance in from day one. What we refuse to claim is part of the pitch. And the same‑day artifact closes every loop: the meeting's output becomes the next meeting's ground.
Where it goes. Each account climbs a ladder. It starts as an instrument on one workflow: decisions priced, predictions sealed, outcomes graded. It grows into a governed capability inside their estate. And it lands at the goal: enterprise deployment of the Boombox capability substrate, activating legacy software as governed capabilities across a network downstream. The legacy products get wrapped once, registered once, and become callable everywhere, with every action carrying the full record: what was known, what was chosen, what we predicted, what happened, and what that taught the system.
Why the network wins. A workflow built in one lane drops into the others, so value compounds across workstreams instead of resetting per project. The partner's own network is the first economic edge: for Mediaocean that is Omnicom through the OMG seat. We are not selling a tool into a stack. We are activating the stack they already own.
We argue. We make the calls together. Every meeting feeds back, and this deck becomes the run book.
This is the machine we are running on ourselves this week, and today is step one. We argue each arc out loud, then we make the calls together: every prediction on boards 03–05 gets a number the whole room signs, sealed before the 5pm call. Each meeting is recorded from minute zero. The transcript comes back to Claude the same day, and this deck re‑rooms: calls graded, open rows filled, next moves drafted. That updated deck is the run book we carry into the next room.
Each block ends with owners and dates, or it did not happen.
| When | Block | Leave with |
|---|---|---|
| 2:00–2:10 | The openerAdam gives the board 00 talk: where we deploy, how we get in, where it goes. No slides voice; this is the point of view. Then the clip on board 01 seals it. | The room aligned on the sandbox and the goal sentence. |
| 2:10–2:40 | The call deck, firstThe 5pm call is on everyone's mind, so it goes first: walk the call deck board by board with one question per board: can we say this at five? Fix anything the room will not sign, in the room, and flip the three switches while everyone is fresh. The deck: /call · approved · adversarially reviewed, structure stood, word‑level fixes applied. | The 5pm deck approved, or edited and redeployed. Switches flipped, owners set. |
| 2:40–3:40 | MediaoceanThe deepest block; Thursday is in person. File on screen, argue the nine questions on board 03, tear at the strawman. | The two opening lines said aloud and owned. The fallback decision class picked. Who carries each Thursday beat: Adam story and politics, Spencer commercial and the paper, Matt method and data rights, Jordan runtime and security. |
| 3:40–4:05 | iHeartFriday is a conversion call: dates for the three agreed proof steps (walkthrough · S3 extract · reverse demo) or nothing. Board 04. | Who briefs Phil and AJ. The extract spec owner. The reverse‑demo shape. |
| 4:05–4:30 | Polymarket, the dry runStill last for freshness, but now pure rehearsal: the deck was approved at 2:10. Argue the seven questions on board 05, then run the first ten minutes of the call for real with the approved deck on screen. | Who says the sample ask. What we say on price. Matt's holdout line for Anthony rehearsed once. |
| 4:30–4:50 | The sealRead the predictions on boards 03–05 aloud. Move any number the room will not sign. | Re‑sealed before the call. |
| 4:50–5:00 | ResetWater, screens, breathe. Join the Polymarket call from 508. | Recorder still running. |
Draft numbers below; the room re‑argues them at 4:30 today and seals what everyone signs. The Thursday transcript grades them Thursday afternoon.
| id | What we say happens in that room | Confidence |
|---|---|---|
| MO‑1 | Bill Wise, talking about where Mediaocean goes next, uses ownership language about the orchestration layer: "own," "control," or "the next phase." | 85 |
| MO‑2 | Iván Markman (President & COO, holds the AI‑transformation mandate) asks the governance question: who runs this system, and where does it sit. | 70 |
| MO‑3 | The creative‑optimization workstream (the test Zvika set up through Alison.ai) is raised by the Mediaocean side before we open it. | 65 |
| MO‑4 | Drew Kane (Prisma CPO) restates his boundary position in some form: publisher decisioning stays within existing publisher and partner stacks. | 60 |
| MO‑5 | We leave 120 Broadway with at least two of the three calendar invitations we came for: a named closed campaign for the retrospective · a runtime and security session with Jordan · an Omnicom working session. | 50 |
The yes exists on tape: "no reason not to plug some campaigns in." Friday turns it into three dated steps: the walkthrough, the S3 log extract, the reverse demo.
The thread: Jul 30 demo call, 32 minutes: Adam, Spencer, Matt, Jordan with Josh Backer and Matt Pilcer. Matt walked the live dashboard, the room engaged, Adam floated the browser‑plugin overlay path, and the yes arrived on tape: "no reason not to plug some campaigns in." Friday's job is conversion: three dates or nothing — the walkthrough, the S3 extract, the reverse demo.
Draft numbers below; the room re‑argues them at 4:30 today and seals what everyone signs. The Friday transcript grades them Friday afternoon.
| id | What we say happens on that call | Confidence |
|---|---|---|
| IH‑1 | Josh Backer re‑raises the natural‑language rules editor (his own ask from the Jul 30 call) without us prompting it. | 80 |
| IH‑2 | Security or NDA review is raised as a gate that must clear before any campaign data moves. | 70 |
| IH‑3 | A named owner for the S3 log extract (Phil or AJ on their side) is committed on the call. | 65 |
| IH‑4 | All three proof steps agreed verbally on Jul 30 get calendar dates by end of day Friday: the campaign walkthrough · the S3 log extract · the reverse demo on their data. | 55 |
The deck is the iHeart playbook translated for a room we hold no data from yet: instead of demoing our instrument, we drew their screens (their columns, their funnel, their World Cup flight) and put the discipline itself on the surfaces: the held row, the sealed holdout, the miss that stays up, closing with a five‑item ask sized under everyone's approval threshold. Its one honest risk: every number on it is invented and the deck is a forwardable URL, so the candor has to stay attached to the pixels; the overnight review re‑anchored every board's labels for exactly that. Reviewed adversarially against the iHeart research (five lenses, then a judge): the structure stood, four word‑level fixes applied, four bigger "improvements" rejected on purpose.
Draft numbers below; this is the first board we seal, since it settles in three hours. Tonight's transcript grades it tonight.
| id | What we say happens on that call | Confidence |
|---|---|---|
| PM‑1 | Peter Muzzonigro (Publicis, our champion, who built this call) vouches for Precise unprompted inside the first ten minutes. | 85 |
| PM‑2 | Anthony Khaiat (Polymarket data scientist, ex‑Kalshi, the technical judge) asks the identification question: how do we separate what the spend caused from what would have happened anyway. | 75 |
| PM‑3 | Marketing‑mix modeling (MMM, the incumbent way to allocate budget) is raised by their side before we mention it. | 70 |
| PM‑4 | They accept, in principle, our five‑item data request for a retrospective on one closed flight (AppsFlyer rows · one DSP's delivery and cost · field dictionary and join keys · first‑deposit definition · their current decision report). | 60 |
| PM‑5 | No delivery date for that data gets fixed live on the call; the date lands in the follow‑up thread instead. | 65 |
Every live surface in the estate. The files are the working instruments and re‑room after every meeting; the research annexes hold the raw depth; the decks are what partners hold.
After each meeting this deck re‑rooms: calls graded, misses shown with the hits.
At 4:30: read the fourteen calls aloud. Sign or move each number. Re‑seal.
Master deck 02 starts where Friday ends.
Computed over the fourteen prediction rows on boards 03–05. Seal v2 rewrote every row to stand on its own (who, what, which meeting); every confidence is unchanged from v1. The last seal stamped before 5:00pm Wednesday is the one that counts.
When it helps, we show a partner the sealed sheet from our own meeting with them: graded, misses included.
Nobody our size walks in this prepared. Nobody their size expects it.
Thursday we walk in four wide. Every question already has a name on it.
Miles did not set the lap record by driving angry. He set it by knowing exactly when to go. This is the lap.
Look at what is already on the table. Three working files that re‑room after every conversation. A product a COO watched run live and said yes to in thirty‑two minutes. Forty‑one numbered questions that turned into three rooms full of people who want to talk to us this week. Fourteen sealed calls we are brave enough to grade in public. Nobody our size walks in this prepared, and nobody their size expects it.