How it works
Each stage is a job with an input contract, an output contract, a spend cap and a record. Nothing runs past a checkpoint until you choose.
- Brief. You fill in ten short fields. A planning model writes a decision card, invariants and 12 proposed directions.
- Exploration. 12 directions on 3 models each, plus 6 single-model wildcards: 42 concepts. Hard spend cap of $5 per exploration.
- Review. A public page with every concept, its prompt, model, cost and time. Filter, search and shortlist.
- Round. Pick up to two anchors, tap what to keep and change, set how far to go. You confirm the contract, then it runs.
- Kit. Later: a constructed vector identity with an automated audit. Not available yet.
Pricing
| What | Now | Later |
|---|---|---|
| Exploration | $0 | $29 |
| Extra round | $0 | $19, or 3 for $49 |
| Private | $0 | $49, free with kit |
| Kit | not yet | $199, exploration price credited |
Generation costs $1 to $3 per project, and prices are set close to that. A failed generation is re-run or refunded automatically. Taste is not refundable.
Prompt skeletons
The planning, exploration and refinement prompt skeletons will be published here once the pipeline is ported from the lab.