Get paid for the electricity you don’t use when the grid is about to fall.
An agent watches Colombia’s real wholesale price, smart meters sign what they measured, and a contract on HSK settles the verified reduction in stablecoin — within minutes, to anyone with a meter. No invoice, no reconciliation, and no human signature anywhere in the payment.
The problem
The cheapest megawatt is the one nobody uses. Nobody gets paid for it.
Every evening between six and nine, Colombian wholesale electricity costs well over half again what it costs the rest of the day — thermal plants set the price when hydro runs short of the peak. The figures above are read live from XM, the system operator, and converted at the official exchange rate. In a dry year the gap widens sharply, which is the risk the country has been managing all through 2026.
The fastest capacity any grid has is demand that simply steps aside for those three hours. There is no way for a household or a small business here to be paid for stepping aside.
Why it doesn’t exist here
Settlement costs more than the energy saved.
Demand response is an established, roughly eight-billion-dollar market in the places that have it. It has not reached Latin America for an unglamorous reason: metering, verifying, contracting and paying ten thousand small participants costs more than the electricity they would save. The economics fail on paperwork, not on physics.
That paperwork is exactly what an agent, a stablecoin and a contract delete. This is the whole thesis, and it is the only reason a blockchain belongs anywhere near this problem.
- Wet signatures & notary contracts
- Manual reconciliation, paid on invoice
- Verification costs more than the energy it confirms
- Every reading carries the device key's signature
- Counterfactual baseline rebuilt by the agent
- Settlement and protocol fee release in one transaction
How it works
Four steps, no human in the loop.
- Step 1The agent senses
It reads XM's published hourly price and reservoir level, compares the evening window against that day's own average, and decides by arithmetic whether the hour is worth an event.
Source: XM REST API - Step 2The meters sign
Every fifteen-minute reading is signed by the device's own key. The contract verifies a meter signature exactly as it verifies a human wallet.
Auth: EIP-712, device key - Step 3The agent verifies
It rebuilds the counterfactual from five ordinary evenings, subtracts what was measured, and compares the result against the reduction the site committed to.
Compute: baseline delta - Step 4The contract pays
If and only if the commitment was met, the escrow releases. Ninety per cent to the site, ten per cent to the protocol — atomically, in one transaction.
Execution: release() on HSK
The contract requires two signatures. One belongs to the meter, one to the agent. Neither can release funds alone, neither can change the amount, and a person who broadcasts the transaction is paying gas rather than approving a payment — their signature is not in it.
The business
Ten per cent of every settlement, enforced inside the contract.
Atomic split
The fee is not an invoice anyone has to chase. It is a split inside release(): the same transaction that pays the site pays the protocol, or neither happens. On top of that sit a per-dispatch fee the agent charges the offtaker machine to machine, and a subscription for committed capacity.
Arbitrage mechanics
The offtaker pays fifteen cents for a kilowatt-hour it would otherwise buy at thirty-one during the evening window. That is roughly half price for the same relief, and it beats a blackout by considerably more. The arbitrage is the business; the contract is only what makes it cheap enough to run at scale.
Beyond electricity
The contract knows nothing about energy.
It knows that an offtaker funded a budget, a device signed a measurement, and an agent verified that measurement against an agreed baseline. Change the sensor and the same machinery pays for cubic metres of water not drawn during a drought, or for verified fire-risk mitigation around a páramo. Electricity is the first vertical because it is the one on fire this month.
What this is not
The honest part.
The agent reads XM’s published hourly spot price and reservoir level, and converts with the official exchange rate. XM publishes a couple of days behind, so the app always shows which day the figure belongs to. If those sources cannot be reached it falls back to a fixture that says so on the page, in those words.
They are synthetic and signed by a development key, because there is no meter connected — and even with one, the baseline needs five days of history before it exists. A signature proves non-repudiation, that this device said this, and nothing more. It does not prove the meter was not physically tampered with; that needs a secure element, and it is the next step rather than a solved problem.
It is agreed in the programme terms, not a measurement of a world that did not happen. No utility has signed anything. What is real is the contract, the verification, and the payment you can watch execute.
The next step
From a simulated meter to a real one.
A real device has to solve three separate problems, and solving one does not solve the others: measure the watt-hours, attest that this specific device said so, and transport the statement out of the building. Everything downstream — verification, the baseline, the payout — is already built and does not change.
- 01Measure
CT clamp + metering IC
- 02Attest
Secure element, secp256k1 signature
- 03Transport
WiFi / LTE-M / LoRaWAN
- 04Settle
Agent verifies, contract pays — unchanged
The middle card is the hard one. The cheap secure element everyone reaches for first signs a different elliptic curve than Ethereum uses — a mismatch that has sunk other projects after the hardware was already ordered. docs/hardware-roadmap.md in the repository writes out the trap and the three honest ways around it, plus a costed, phased pilot starting at one cooperating site for about USD 150 in hardware.