The **solver workspace** is where paid accounts manage the equilibrium strategies they own: the equilibriums themselves, the solvers that serve them, the learning races that train them, and the abstractions that make large games tractable. It lives at [mieza.ai/gto/workspace/training](https://mieza.ai/gto/workspace/training) and is reached from the main navigation under **Strategy → Solver workspace**.

The workspace is scoped to whatever you have selected as your active account: your personal account, or one of your organizations. Switching organizations switches the inventory you see.

## Who Can Use It

The workspace requires a **Pro** or **Elite** entitlement. Free accounts that open it are sent to the billing section of Settings. Platform administrators have their own management pages and are not affected by this rule.

Within an organization, what you can do depends on your role:

| Capability | Organization owner / admin | Organization member | Personal workspace owner |
|---|---|---|---|
| Browse equilibriums, solvers, learning races, abstractions | Yes | Yes | Yes |
| Create an equilibrium | Yes | — | Yes |
| Save a training configuration | Yes | — | Yes |
| Enable or disable an owned solver | Yes | — | Yes |

Management actions also require that the resource belongs to the selected scope. An organization admin cannot manage a solver owned by a different organization, and a personal workspace cannot manage organization-owned resources.

## Workspace Sections

The landing page links to four sections. Each mirrors a concept documented elsewhere; the workspace is where you act on them.

### Equilibriums

An [equilibrium](/docs/what-is-an-equilibrium) is a computed strategy for a specific game configuration. The equilibriums list shows the equilibriums you manage, filterable by status.

**Create Equilibrium** opens a form where you choose the game (its [strategic context](/docs/what-is-a-strategic-context)), name the equilibrium, set its game parameters, and decide whether it is public. Public equilibriums are visible to all users, so avoid public visibility for proprietary business models or confidential pricing assumptions.

An equilibrium's detail page has three tabs:

- **Overview** — the equilibrium's game, owner, and training status
- **Solvers** — the solvers that load this equilibrium (see below)
- **Settings** — the training configuration

### Training Configuration

The Settings tab holds the default CFR parameters that any future training run of this equilibrium will use. The fields are:

- **Total Iterations** — how many CFR iterations to train (on the order of 100,000 for Kuhn poker, millions for Texas Hold'em)
- **Checkpoint Interval** — iterations between strategy saves and evaluations
- **Discount Interval** — iterations between regret discounting passes
- **Strategy Update Interval** — iterations between updates of the average strategy
- **Prune Threshold** — iterations before pruning starts
- **Discount Threshold** — iterations before discounting starts
- **Parallel Traversals** — run CFR tree traversals in parallel for faster training
- **Ephemeral Mode** — skip disk persistence (for testing only)

**Save Configuration** stores these values on the equilibrium. Saving is disabled while a learning race already exists for the equilibrium, so a configuration cannot change underneath a run in progress. See [What is a Learning Race?](/docs/what-is-a-learning-race) for what each parameter does during training.

### Solvers

A [solver](/docs/what-is-a-solver) loads an equilibrium's strategy into memory so it can answer queries. The solvers list shows every solver you can see, with its equilibrium and current state.

If you manage both the solver and its equilibrium, the solver's detail page shows an **Enable** or **Disable** button. Activation is asynchronous, and the page distinguishes two states:

- **Desired state** — the badge shows *Enabled* or *Disabled*: what you asked for
- **Actual state** — while the runtime is still catching up, an amber badge shows *Starting...* or *Stopping...*

After you request a change, the page checks status automatically, backing off from a few seconds between checks to half a minute. After eight checks without convergence, automatic updates pause and the page tells you to refresh manually. A failed request keeps the last confirmed status on screen and explains what went wrong.

Two constraints apply at the platform level. A solver can only be enabled once its equilibrium has a **published** strategy artifact, and it cannot be enabled while a learning race for that equilibrium is queued or running. The reverse also holds: training cannot start while any solver for that equilibrium is enabled. Disable the solver first.

### Learning Races

The [learning races](/docs/what-is-a-learning-race) list shows the training runs and prepared configurations visible in your scope. A race's detail page shows its status and its **event log**, the chronological record of what the trainer did, which updates while a race is running.

### Abstractions

[Abstractions](/docs/what-is-abstraction) are the game-tree compression pipelines that make large games solvable. In the workspace they are read-only: you can browse the abstractions available for your games and open their details, but creating or deleting abstractions is not available to customer accounts.

## What the Workspace Does Not Do Yet

The workspace page says it plainly: **starting training is not available here yet**. Today the workspace lets you prepare everything a run needs (the equilibrium and its training configuration) and lets you activate solvers whose strategies have already been trained and published. Launching a learning race from the workspace is a separate capability that is still in progress.

## See also
- [What is an Equilibrium?](/docs/what-is-an-equilibrium) — the strategies the workspace manages
- [What is a Solver?](/docs/what-is-a-solver) — enabled and disabled states
- [What is a Learning Race?](/docs/what-is-a-learning-race) — the CFR training runs configured here
- [What is Abstraction?](/docs/what-is-abstraction) — game-tree compression
- [Organizations](/docs/organizations) — roles and organization scope
- [Plans and Billing](/docs/plans-and-billing) — Pro and Elite entitlements