Built on Workbench

Here's a thing worth knowing early: the AI in every ZeroWidth product is a Workbench flow. These are the same kind of flows, running on the same engine you use to build your own.

• zv1, the assistant, is a tool-using agent flow.
• Compass extracts pages from your documents and runs stakeholder interviews with flows.
• Caliper scores your evals with an LLM-judge flow, and generates datasets with another.
• Prism runs its idea permutations, desk research, adaptive interviews, and study analysis on flows, too.
• Ledger and Napkin reach zv1 the same way — the scribe that records decisions and the eyes that read a sketch are flows.

When Caliper judges an output or zv1 drafts a flow for you, that's a flow run — the exact same path as running one of your own.

Why it's built this way

• Transparency. The AI you're trusting is built with the tool we hand you, and there's nothing in it you couldn't build yourself.
• Proof. It's the strongest evidence that Workbench is production-grade. Our own products run on the same engine you get.
• One loop for everything. Because they're flows, our features live in the same improvement loop as yours: versioned, measured in Caliper, improved. We eat our own cooking.

So when you build a flow, you're using exactly what runs ZeroWidth.