Discover
Autonomous multi-file evolution
Describe a goal and the agent runs the full cycle autonomously — decomposing it into Ideas, building Experiments, generating multi-file Versions, and carrying only the best forward into the next generation. Exploration and combination in one run.
Need per-file diff exploration? Targets · Need to combine specific changes? Optimise
Starting a Discovery Run
From the Discover page (accessed via the Optimise tab), describe what you want the agent to discover or optimise in the input field, then submit your goal.
Starter prompts
Not sure where to start? Use one of the built-in prompts:
- Optimise performance
- Hardware-aware optimisation
- Improve model accuracy
- Cheaper, faster agent
- Find and fix the highest-value bug
- Cut build or CI time
Configuration
Before starting a run, configure how Discover explores your codebase:
| Setting | Description |
|---|---|
| Model | The AI model used for the discovery (e.g. GPT-5) |
| Generations | Number of evolutionary generations to run (default: 3) |
| Versions per gen | Number of versions generated per generation (default: 5) |
| Run code after each version | Executes and validates each version as it is generated |
| Manual approval mode | Pauses between generations for you to review before continuing |
The total number of versions created is Generations × Versions per gen (e.g. 3 × 5 = 15 versions).
Discovery Runs Panel
The left panel lists all your Discovery runs for the current project. Each run shows its status and goal. When no runs exist yet, kick off your first run from the form on the right.
How Discover Works
Each Discover run follows an autonomous evolutionary cycle:
- Ideas — the agent decomposes your goal into candidate approaches
- Experiments — approaches are built into concrete code experiments across multiple files
- Versions — each experiment produces one or more code versions
- Next generation — only the best-performing versions are carried forward; the cycle repeats for the configured number of generations
This means later generations are built on top of the best earlier results, progressively improving the solution.
Next Steps
- Code Targeting — identify specific targets for manual, per-file exploration
- Creating an Optimisation — combine specific versions you've already selected
- Reviewing Results — understand the output metrics