AI × Open Source in Energy Markets
From hidden models to shared assumptions
It’s becoming harder to think about the energy transition without also thinking about open-source and generative AI.
The opportunities are not just technological. They are structural. Generative AI becomes more powerful when it is paired with open data and open implementations, when the underlying logic can be inspected, adapted, and extended.
To understand what that might look like in practice, it’s useful to look at a field that is intrinsically based on openness: mathematics.
In conversation with two mathematicians recently, I was struck by their optimism and curiosity about the potential for AI in their fields.
Two concerns came up repeatedly. First, that the craft of mathematics should remain intact. Second, that students should still be able to explain the mathematics they develop with AI.
Craft and explainability: good guiding principles.
Mathematics is in many ways already open-source. Progress tends to be highly visible, cumulative, and contestable.
It is no surprise then that there are already public repositories tracking AI-supported progress for the mathematical community.
It is difficult not to contrast this with parts of the energy industry.
We rely heavily on models—of grids, of markets, of asset behaviour—but we rarely develop them in a shared space. The assumptions sit within companies, consultancies, or system operators, each reflecting a slightly different internal logic. When concerns arise, the models themselves are not always accessible.
So the conversation takes on a particular shape. We regulate against outcomes we are concerned about, but cannot always examine in full.
Stratnergy has been moving, slowly, in a different direction.
Earlier work on “true indices” focused on the idea that a meaningful benchmark should be transparent, systematic, and replicable. In practice, that points naturally toward open implementations—toward strategies that can be inspected and rebuilt.
More recently, I’ve started to collect examples of open-source work that has relevance or potential across trading strategies, forecasting approaches, and grid models. A first version of that sits in the subscriber terminal as an evolving OSS index and survey.
These ideas are not only about building tools or indices. They begin to touch the way we reason about system risk.
A lot of current discussion in power markets is shaped by worst-case thinking. Batteries might synchronise. Flexibility might concentrate. Trading logic might, under certain conditions, amplify stress at the boundaries of 15-minute products.
These concerns are not unreasonable. But they are difficult to examine directly when the models behind them are not shared.
Which raises a more general question.
Why not negotiate these fears through shared, open models?
One could imagine a layered approach. Traders contributing stylised strategies. System operators contributing constraint logic. Different actors making their assumptions explicit, not to eliminate disagreement, but to make it more precise.
In that sense, open-source becomes a way of structuring and developing the conversation.
The interactive version of this newsletter at stratnergy.online began, in part, as a repository of index strategies. I stepped back from that to work on a stronger foundation. It now feels possible to return to advanced features. I’m particularly interested in the potential for open trading strategies, grid models, market forecasts, asset virtualisation logic, etc.
If some of the tensions in the system are rooted in models we cannot fully see, then opening parts of that modelling process may change the discussion itself. Not by removing risk, but by making it something that can be examined, challenged, and refined collectively.
That, at least, seems worth exploring.