The Venue

The Venue

Palaestra is live at palaestra.thrivetech.ai.

One line

Palaestra is the venue where people, and later agents, learn and compete at the interface between the electric grid and very large loads, by playing as the large load.

The thesis

The grid and large-load interface is the live question in power systems. Interconnection queues, ride-through obligations, load flexibility, reserve impacts: practice is outrunning the standards that govern it.

The bodies that write standards find gaps by literature review and expert judgment. That is slow, and it is biased toward what experts already believe.

Palaestra finds gaps by experiment. Put a player in the data-center seat at the grid interface. Give them real control points. Score them on tensions that actually tension. Then watch. Every ambiguity a player exploits, every rule that produces a perverse outcome, every place two reasonable players read an obligation differently, is a documented standards gap found reproducibly and cheaply.

Palaestra is a standards-needs discovery instrument disguised as a competition. Nobody attends a workshop called "interconnection obligation ambiguity survey." Everybody wants to beat the scenario.

The seat

The player is a data-center owner and operator. That choice is settled, for three reasons.

It is the fresh seat. Utility-side operations and planning games exist. The load-side seat, negotiating interconnection, choosing ride-through postures, building on-site generation, timing curtailment, has no incumbent.

It avoids a fight. A utility-seat product competes with large grid-software vendors who would work to discredit it. The load seat threatens nobody's business model, and it educates their customers.

It is where the teaching is. The power-systems audience has the least intuition for the load side. The data-center audience has the least intuition for the grid side. One seat teaches both, from the side each audience does not know.

There is no playable grid seat, and there is not expected to be one. Grid changes enter the game as player-initiated requests adjudicated by the scenario, never by a human opponent.

The phases

Phase one: humans, as education. People play, in workshops, tutorials, and classrooms. Every session also produces an episode log, which is the first standards-needs data. This is where the venue is now.

Phase two: agents, as competition. Teams field agents that implement the same competitor contract humans played through. The organizer runs the physics and the competitors run the agents. The boundary between contestant and physics is what gives reproducible scoring and resistance to cheating.

Phase three: policy. Players influence the rules themselves. The game about playing under rules becomes a game about changing them, which is what the real players do.

The ordering carries weight. Humans first means the venue earns its educational credibility before any agent hype, and it means the contract is proven usable by people before machines optimize against it.

The commitment that shapes everything

A human at a browser and an agent at an API are two implementations of the same interface. Phase two is a swap, not a rewrite. That single commitment is why the contract is a document rather than an implementation detail, why the UI is a client with no privileges, and why scenarios are data instead of code.

See The Competitor Contract.

What Palaestra is not

• Not an event. The competition is an occasion; the venue persists.
• Not a scenario library. Scenarios are data the venue loads.
• Not a utility-seat simulator, a market-clearing engine, or a planning tool.
• Not a certified study platform. The physics is believable, and the venue says where the model stops.
• Not a vendor product. Noncommercial and open by default.