Dreamineering

Thoughts · Proof

Ownership moves. The bottle stays put.

The question. When agents call the shots, what decides where the money goes?

Four parts: the question, what we think now, what we’re unsure of, and what would change our mind. It stays open.

People buy stories. Agents can’t.

An agent spending your money needs something it can check. In the digital world that is hard enough. In the physical world it is harder. A record can say anything. A bottle is either on the shelf or it isn’t.

So what counts as proof when the thing is real?

Four questions every physical deal must answer

  • What is it? Identity. This bottle, this barrel, this machine.
  • Where is it, exactly? Position. Not roughly. Exactly.
  • What state is it in? Condition, over time.
  • Who checked? Custody. A named party who carries the cost if they’re wrong.

Miss one and the deal rests on trust you can’t inspect.

One retail case that puts it together

BAXUS trades collectible wine and spirits. It is worth studying because it climbs every rung in one product.

  • It starts with a free tool. Collectors scan bottles in stores and log the shelf price. Each log earns points, and the prices feed a shared database (Pluang). People help each other answer a question they already ask: is this a fair price?
  • Then a vault. Bottles sit in insured, climate-controlled storage. Each one is tied to a digital record, and owners can sell, or borrow against what they hold (Helius).

Ownership moves. The bottle stays put. Someone accountable checks it is there.

Notice what didn’t disappear: the middle. The vault, the checker and the marketplace are still there, and they still charge. What changed is that their job is visible. That is the honest version of “cutting out the middlemen”. Opaque middlemen become accountable services.

Where, exactly?

Ordinary satellite positioning is good to a few metres. For a person with a phone, fine. For a drone landing on a deck, a tractor holding a row, or a robot finding a shelf, three metres is the whole problem.

GEODNET is one answer. Reference stations, run by many owners, measure local satellite errors. The network turns those into corrections, and a machine uses them to fix its position to centimetres, the company says (GEODNET). Station owners earn tokens for taking part.

Precise position isn’t a nice extra. It is evidence. “The inspection happened here, at this time” is a claim a buyer can check.

Eyes at the edge

375ai puts sensor nodes in public spaces. Each node turns video, sound and other signals into counts, classifications and movement before the data leaves the device. It calls itself “an ingredient, not an application” (375ai).

That is the right humility. A read is not proof by itself. It becomes proof when someone accountable uses it to make a decision, and stands behind the result.

A shared field

Machines from different owners can only play one game if they share the markings. The Intercognitive Foundation groups companies working on that, including GEODNET, around nine pillars: identity, fees, maps, sensors, positioning, compute, connectivity, orchestration and standards (Intercognitive).

It is early. Shared markings are a promise of a field, not yet a game.

Two loops, one technology

Game mechanics decide what people do with all this.

Loop one pays the observation. Log a real price, at a real store, and the next buyer decides better. The reward follows work someone else can use.

Loop two pays the pull. Collectibles markets also sell random draws with an instant buyback at a share of the item’s stated value. A buyback of stated value is not a refund of what you paid. Pull again and again, and money leaks away.

Same vault. Same tokens. A different game. One builds a shared picture. The other builds a habit.

What we think now

When agents decide, money should follow evidence of a finished outcome, not the confidence of a claim.

A token records who said what. It doesn’t make it true. Truth comes from the four questions, answered by parties who carry a cost when they’re wrong.

Put it together and the unit of trade changes. Not a product. Not access. A result: this job, in this place, by this time, within these limits, with evidence a buyer can check before money moves. That is our working idea. We haven’t tested it with a customer yet.

What we’re unsure of

  • Who checks the checker? A vault or a sensor network can fail, or lie. What makes the custodian accountable?
  • Can a physical proof be faked cheaply? If it can, rewards will go to the fakers.
  • Who carries the loss? When a machine acting for an agent damages something, the law has not settled whose problem it is.
  • Will agents read the evidence, or just the price? Cheap and unproven may still win.

What would change our mind

  • A lasting physical market where money moved safely on records alone, with no one accountable for the thing.
  • Evidence that buyers, people or agents, consistently pay less attention to proof as it gets cheaper to supply.
  • A reward for observations that held up at scale without being gamed or faked.

Your turn

Pick one thing you would buy, sell or lend against without seeing it. Ask:

  • What is it, and how do you know it’s that one?
  • Where is it, exactly?
  • What state is it in, and since when?
  • Who checked, and what happens to them if they’re wrong?

If nobody can show you, you are buying a story.

Read it with Coordination: keep the shape, win the edges, on devices as players, and Tokenomics: pay the work, not the gate, on who a reward really pays.

Published 5 November 2026. Companies are named as examples of patterns, not endorsements. A reflection on how trade is changing, not financial or legal advice.

Think we’ve got this wrong?

Good. Tell Nav what we missed. The best challenges change the page.