The order book, arriving from the bottom.

People ask when Precise started, and the honest answer is a decade before the incorporation papers. In 2016 Tom Bollich and I founded MadHive to build television advertising software. By 2017 we were building MAD Network underneath it, a blockchain protocol for ad tech, and the way I have always told it, the whole idea was an order book for the internet. A place where contribution gets a price, the price gets tested against a real outcome, and settlement is arithmetic anyone can re-run. Advertising was the first market because advertising is where value attribution is most broken and best paid.

We were early, which is a polite word for wrong about timing. But we shipped. A working proof of concept in 2018. AdLedger the same year, a standards consortium we founded with IBM and TEGNA, because I believed then what I believe now: you own a standard by giving it away. In 2020 MadNetwork ran as a permissioned layer anchored to Ethereum, and AdLedger used it for the first blockchain-based OTT ad deployment, real spots and a real fraud problem, with Beachfront. In 2022 the chain forked into AliceNet. Matt Barlin was AliceNet's CTO. He is Precise's Chief Data Scientist. And AliceNet is more machinery than people remember. Its current generation is live on Ethereum mainnet today, one chain under both companies, and through the Valence identity layer it carries verifiable credentials for people, assets, and organizations, and selective-disclosure proofs: prove the claim true without revealing the data behind it. The order book was always the most important thing we planned on top of that mechanism, before Precise and before Konstant. The blockchain work is also where the media application finally clicked, at the base primitive level, and Grant is the one I thank for that. The market moved on from the word blockchain. The question never moved at all.

The part of it I remember best is not a launch. It was COVID, spring probably, and I was walking circles in my backyard on the phone with Matt, explaining the order book for the internet. By the end of that walk we had agreed on what we wanted the world to look like. Then we went and built the chain, and we got it done: contribution value, credentials, Merkle trees, the ways to catalog across networks. The primitives were real. The world they were built for had not arrived yet.

Three questions, actually, and we have been chasing them the whole time. Who created value. Where should the next dollar go. What does the evidence deserve to claim. Years of contribution research later, the answer has a shape, and it cuts both ways. The primitive is contribution value in the context of an outcome: decompose the inputs into what each piece earned, and decompose the outcome itself into its pieces, the whole chain downstream. The primitive deepens with data, and here is the part I keep turning over: an outcome is never one event. Between the spend and the deposit sits the whole product, onboarding, discovery, the ticket, the notification, dozens of variables downstream of the ad. Decompose the outcome into that chain and every piece becomes another place to score contribution. More pieces, more data, better contribution value. It goes all the way down, which is exactly what the Doors SDK is for: the app SDK we are pitching Polymarket, making bets on the outcome side and carrying evidence back. And the order of operations matters: the analysis comes first, then the bet on top of it. Change these weights, and I bet that happens. That is the call, and it stays a live bet, adjusted as evidence lands. The unit of record is the Call: the bet the analysis decides, written down at the moment it is made. What was known, what we chose, what we predicted at what confidence; then what happened, what contribution the comparison supports, and what it taught. Writing it down when we make it is what keeps the scoring honest. Nothing ships without a call. And the system earns belief in a strange way, by refusing. When it says it has no edge, that refusal is the instrument working. I trust a model that can say no, and so do CFOs.

So here is what Precise is. We build best-in-class Decisioning Economics software, ad tech first, deployed to customers white-label on the Konstant platform. Each customer ends up owning a decision-learning network of their own. Their data stays theirs, their calls stay theirs, nothing pools, and authority never crosses company lines. Only arithmetic does. We are the referee, so we do not take a cut of what we measure. It takes two companies on purpose, because the rails only work if they are nobody's product. The ones who join are our customers. They build their own capabilities on their own context, send them downstream through their own networks, and replace the vendors charging them rent today, one seat at a time: the vampire attack, run by the customer. Konstant's independence keeps those networks theirs. And the dependency runs every direction; this is the case I made when I asked for the room to build it. None of the pieces works as well alone. Without Precise, Konstant stops at organizational intelligence: context without the capability that gives the network gravity. Without Konstant and Boombox, Precise stays powerful but hard to move across companies; without the network to think in, we would have ended up building dashboards. Without the capability network, the blockchain work lacks the agentic economy it was built to support. And the economic alignment does not come from my equity. It is structural: the network consumes the primitive everywhere it runs, so as the network wins, Precise wins. I could not prove that yet, so I proved it the only way available, with my own money and my Konstant equity.

Konstant's own eighteen months went into the network, and into selection, with its own algorithms and patents. Selection is the quiet hard problem here, and there are two selectors, split exactly along the company line. Konstant routes the organization: which capability, which provider, which composition fits its context, authority, relationships, custody, timing, and commercial terms. Choosing context is an algorithm, and TikTok is the proof the machinery works: underneath the feed it decomposes every video into the attributes that predict you watching more. That is contribution analysis running inside selection. Konstant points the same machinery at decisions: select the context that matters for the call, and feed it to the ablation system. Precise runs the evidence-bound court inside the decision: freeze the eligible models, methods, strategies, policies, and surfaces, apply the eval law and the floors, and return a winner, a null, or a refusal. Neither selector silently acquires the other's authority. The score teaches both. And the framework is clean: Konstant consumes the Precise primitive as a licensee, and everything it needs is available now. We just have to get it going. The context provider and the decision primitive compose into one system that stays two companies.

When someone asks how long this took, be careful with the numbers. The eighteen months is Konstant's alone. The research behind the primitive is nine years old, and the people are the proof: Matt and I built the chain that carried it, and he is our Chief Data Scientist today.

We are building this bottom-up, on purpose. Mediaocean, Polymarket, iHeart. Each one buys the next-dollar decision, never the endgame. Each answered question grants access, data, or a relationship, and the system assembles from asked-and-answered rather than from slides. That is the only honest way to build it: customer by customer, call by call, until the graph has earned its way into existence.

Someone will ask about the hyperscalers, so let me answer it here. One of them will eventually build a version of this and make it free, and it will not matter, because we have run this experiment before. The majors, ABC, NBC, the companies whose whole business is their media, never handed their ad inventory and workflows to Google. That is why an independent ad market exists, why ad tech exists, why MadHive exists. Google, Amazon, and Apple could not vampire-attack those estates, so they built their own media companies instead: YouTube, Prime Video, Apple TV+. Company towns. AI brings the same fork to every industry, and nobody wants to live in the company town. Agents coming online with crypto make the online world bidirectional, the way the real world already is: any party can buy, sell, and settle with any other, instead of just consuming what the platforms serve. The alternative has to exist, and we are building it.

And then the endgame, which I will say plainly in this room. Work is becoming multi-agentic. When agents do the work, contribution has to be valued and settled machine to machine, at a grain no invoice can reach. Precise values the contribution. The blockchain floor underneath, the AliceNet lineage, settles it. The first rung already runs in our code today, signed verifiable credentials and Merkle proofs in CI; a first anchor path already works at dev tier on the testnet; and everything in between refuses to claim what it cannot yet perform. That refusal is the house style. That is the order book, arriving the only way it ever could, from the bottom, one customer and one call at a time. The chain gave us the floor. The decision software is the primitive that goes with it. And Konstant lets it proliferate into AI-native work, which is where the new internet will actually be. We held this idea through a whole hype cycle and its hangover. Now the timing is ours. One more thing, and it matters as much as any of it: this only works as a shared point of view. Argued, improved, and owned by everyone in the room, with shared leadership behind it. Today is for buying in, or for making it better. Go like hell.