First Look: Direct Trade Protocol (DTP)
Agentic on-chain commerce for the B2B wholesale of physical goods
Something I’m Noodling On (and I’d like your take)
Thirteen years ago, Erin and I started Yellowbird Foods in our kitchen. We made a hot sauce people liked, and eventually it ended up on shelves at over 17,000 retail doors.
Along the way, I accidentally got a detailed education in how the food supply chain actually works (and doesn’t). Brokers. Distributors. Slotting. Net-30 to Net-60 payment terms. Fees, deductions, lost freight. Extractive relationships everywhere you turn. The coordination overhead (spreadsheets, phone calls, faxes) is staggering for what actually gets accomplished.
None of this is new. Everyone in food knows it. The system is inefficient, expensive, and doesn’t do many favors for the producers at the start of the chain or the consumers at the end. You’ve gotta be well-funded to even make it out the gate. Even the companies in the middle can barely figure out how to turn a profit.
A USDA estimate puts the U.S. food supply chain’s middleman and coordination overhead at 20-30% of total value. Somewhere in the range of $375–750 billion annually. That’s a very large tax for a lot of paper-pushing.
I’m 1000% sure there’s a better way and I’ve been writing about it and noodling on it for a few years.
An Important Preface
I’m not about to suggest “AI can do this better”. It’s 2026 and we’re all sick of hearing that as a blanket statement.
What I’m about to suggest is that we actually use modern technology to make our industry and our lives significantly better.
The CPG Food industry seems to always be the last person still ordering from the Sears catalog when the rest of the world has the internet.
The Idea
What if the coordination layer itself wasn’t a corporation or a piece of expensive software? What if it was just a protocol? Open rules that any buyer, seller, or AI agent could implement, with settlement handled by a smart contract.
What if any buyer could work with any seller and we could knock a significant amount of friction and extraction out of the middle? What if that value could return to the system and make the system more valuable instead of enriching a few big companies?
My working title for this is Direct Trade Protocol (DTP).
The basic concept: sellers (or their agents) post what they have, buyers (or their agents) post what they need, a matching engine surfaces fits on both sides, and a smart contract handles escrow and releases funds when delivery is confirmed. No broker in the middle. Reputation lives on-chain and travels with the participant, not with a specific platform. Shopify wouldn’t own your sales data. NetSuite wouldn’t own your financials.
A DeFi layer lives in the middle and more than halves the cost of payment term capital while distributing the earnings to anyone staking funds for yield.
A freight layer also lives in the middle, matching supply and demand in real time.
Entities are real, certified, and reputable and their reputations, transaction histories, and validated certifications travel with them wherever they take their accounts and their catalogs of goods.
Think of it the way HTTP works for the web. HTTP doesn’t own the internet — it just defines how computers exchange information, and anyone can implement it. DTP attempts to do the same thing for wholesale trade of physical goods.
The repo is here: github.com/runquik/direct-trade-protocol
Where this fits in the current landscape
Let me be straight: I’m not working in a vacuum.
Tokenization of physical goods is a real and growing space. Projects like VeChain and IBM Food Trust are doing important work on provenance and traceability. OpenAI and Stripe just released the Agentic Commerce Protocol. Google and Shopify released the Universal Commerce Protocol. Stripe is building a new blockchain specifically for agent-native transactions.
The thing I notice is that most of this work is either focused on tracking goods (not trading them), or it’s aimed at B2C retail (not B2B wholesale). The piece that seems underbuilt is a coordination protocol specifically for wholesale physical goods. One that handles the full arc from intent to matched offer to contract to settlement, without a platform extracting rent in the middle.
That’s the gap I’m trying to fill. But I hold that view loosely. I’m likely missing something.
Where DTP is today
Pre-alpha. Maybe 10% of the way to something you could use in production.
The spec is written. The smart contracts compile. The identity model, matching logic, finance layer, and freight integration are designed and partially built.
Yeah, I did a lot of it with AI. I’m not a developer and am not going to pretend to be. What I am is a human who has lived in this industry for 13 years with his eyes wide open to who the current system serves.
There’s a long way to go but I think it’s valuable and I think it serves more people if it’s open source: if neither myself nor any one person or investor group stands to make a billion dollars on its success. I also think these tools are going to get built one way or the other, and we can choose right now if we want to be customers/users or owners.
The whole point is to make this public infrastructure — not a product with a revenue model, not a company with an exit. If this works, the value goes back to the people in the supply chain, not to whoever owns the platform, because we all own the platform.
A genuine ask for help
A few things I’d love to hear from people who know this space:
Is this gap real, or is something already doing this well that I’m not aware of?
For food operators: does the description of the coordination overhead match your experience? What’s the actual friction that matters most?
For builders: what’s the first thing you’d want to see get built?
For anyone: what’s the biggest flaw in the premise?
The GitHub is there if you want to look under the hood. And I’m genuinely interested in what people think.
It gives me some anxiety to post this in public because I’ve actually spent a while trying to make v0 make enough sense to show. I hope this concept finds some purchase with people who are as fired up as I am to shake the industry up a bit. People who think an industry devoted to feeding people can be less extractive and more human.
Sincerely,
George
Hey George, I've thought about this too. I appreciate you putting more eloquent words to it than I have been able to.
The coordination overhead is definitely real. We were smaller than your 17k doors, but had separate logins to: KeHE, UNFI, each individual retailer site (Walmart, Kroger, Whole Foods, Target, CVS and many more smaller chains), SPINS data, WERCsmart, IX-One global, Bazaarvoice and more. Working with each of these tools was like working with software from the early 2000s, and was a full time job for at least one person. Not to mention each one decides to email you nonstop, and you have to sort through figuring out which emails matter and which ones are irrelevant to you.
I do think AI is at a place now where a lot of the admin overhead of running across all these portals can be reduced, especially if we keep pressure on them to open up their APIs. But to your point, there's got to be a better way overall.
My hunch is that the solution lies in figuring out what value each player in the chain provides to the retailer.
e.g. what value does the distributor offer to the retailer? Fewer trucks unloading at the docks each day? More reliable inventory supply because you know the distributor will keep a nearby warehouse in stock, so it's easier for you to do the same? Is there more to it than that? How can DTP help solve some of those problems for the retailer?
If you can go through the chain and figure out the value add to the end retailer for each one and building around that I think that's where you can start really chipping away at the status quo (while also improving it for the little guys - us!)
Hi George, huge fan of everything you have built and of your substack. My two cents (from my economist perspective): the challenge is not what you are building, which is deeply needed, or the tech, which is almost trivial. The challenge is that what you are proposing is a two-sided marketplace (you have to bring in both supply and demand, so Facebook), and that is about one of the most difficult things to build, which is why we are all still on Facebook and linkedin, even though most people hate it, because the network effects are so strong. The problem in your idea is incentives, as in why would supply side (producers/manufacturers) have an incentive getting here when brokers do the search part for them? and then you have brokers and distributors who have an incentive in making sure this fails. And then you have demand side, which is fleeting because so many people enter the food industry full of passion but low in knowledge and run through their money faster than they can learn what works and doesnt. So I think the problem to build this is not the tech, but figuring out how to align incentives for all players so that they help you build it themselves (self builds), ie the human side is the problem. To my knowledge Canadian Food Innovation Network has been trying to build this for a while, but usage is low.