What if machines could learn from each other’s experience?

Imagine a battery storage site in the third week of a heatwave. Its dispatch software recommends selling power into the afternoon price peak. The operations team has been watching the battery for days. Under sustained heat, it has been delivering less usable energy than the software expects. The team holds some power back. When demand peaks again that evening, the site still has its reserve.

A year later, in another city, an operator running the same type of battery faces a similar decision. The first team learned something worth knowing. The second team has no practical way to benefit from it.

We have learned how to connect machines across the world. A harder question follows: how can the experience of operating one machine become a better starting point for the next?

A manufacturer’s manual gives the battery’s rated capacity. A sensor gives its temperature. A dispatch model recommends when to discharge. The most useful record brings those pieces together and follows the story one step further: the operator overrode the recommendation, the controller executed a different schedule, and the meter recorded what the battery actually delivered.

The distinction matters. If the override goes unrecorded, a model may learn that its original recommendation worked. A trusted record preserves enough context to tell the difference. For TLAY, that is the value of trusted IoT data: being able to trace an observation to its source, distinguish a proposed action from an actual one, and test a prediction against a measured result.

Across energy, mobility and manufacturing, businesses acquire knowledge this way every day. Much of it remains inside a site, a service contract or an experienced team. A company can buy more computing power. It cannot buy a second copy of a summer that only one fleet has lived through.

A company can buy more computing power. It cannot buy a second copy of a summer that only one fleet has lived through.

Put yourself in that operator’s position. You might gladly help the industry understand how batteries respond to heat. You would have little reason to give competitors your bidding strategy, reveal a customer’s contract, or hand a platform an unrestricted right to sell what it learns from your work. The modelling team across the table faces a similar choice: it wants access to better operating evidence while protecting the model it spent years developing.

There is an intellectual precedent for this ambition. In its 2015 founding statement, OpenAI called for research to be shared, institutions to collaborate and the benefits of AI to reach people broadly. Industrial AI gives that aspiration a concrete test: can the enterprises whose experience makes a model useful have a lasting say in its use and a stake in its success?

We believe they should. An operator ought to be able to contribute a lesson without surrendering its business. A model builder ought to be able to collaborate without giving away its weights. Both should know, before they begin, what others may learn, who may use the result, and what happens if a customer pays for it.

TLAY’s long-term ambition is to build shared infrastructure through which industries can develop an IoT Frontier Model together. We would open the protocols for participation and verification. Members could jointly govern the capabilities they build and agree on ownership, access and revenue rights for each shared model. Individual datasets, proprietary models and customer relationships would remain under their respective owners’ control.

A battery project makes this concrete. An operator authorizes use of selected heatwave records for a defined forecasting task. A model team licenses its model for that task. Confidential computing could allow both assets to be used in an agreed, verifiable environment, with limits on what may leave it. An independent site then tests whether the resulting model forecasts usable capacity more accurately than the existing service and a model trained by one operator alone. That result would tell us more about the model’s frontier than its size or its node count.

If the improvement holds, member-operated nodes could deliver the validated service. The node that serves a customer should be paid for that work. The operator whose rare records made the improvement possible should also be able to earn an agreed share, even if it never runs a node. The model developer has a claim for its accepted contribution; TLAY can earn fees for deployment and operation. Blockchain records can let each participant check which version was used, what revenue was received and how it was divided, without making the underlying records public. No single operator should be able to rewrite those terms after the work has been done.

That result would tell us more about the model’s frontier than its size or its node count.

The first customer need not wait for a frontier model. Through the TLAY AI Platform, a site could connect its battery to an existing forecasting and optimization service, retain local control over safety and reserve limits, and pay for dispatch support under agreed terms. That gives the customer a useful service today and establishes a baseline against which a future model must earn its place.

One dispatch can then matter in three different ways. Its delivery record can support machine payments by showing what service should be settled. With separate permission, a history of performance and cash flows could help a financing partner assess the asset for machine finance. With authorization for learning, the difference between forecast and outcome becomes evidence for the IoT Frontier Model. Each use answers a different question. Together they explain why trust, intelligence and commerce belong in the same architecture.

Now return to the second summer. The operator in the other city receives a forecast informed by performance under sustained heat. Its team tests the forecast against local conditions and chooses how much energy to retain. It has never seen the first operator’s customer list or trading strategy. Yet a lesson from that first site has travelled. If the second site pays for the improved service, the first can verify its share.

That is the scale of the opportunity we see. It begins with two sites and one decision. If the arrangement works, a manufacturer has a reason to bring a new generation of cells into the network. An operator in a different climate has a reason to contribute the case the model still gets wrong. Each addition can make a specific capability better, and each improvement can be tested in the world where it must work.

Our ambition for TLAY is that the next machine can begin where another left off—and that the people who made the lesson possible retain a stake in what it becomes worth.

10 November 2026