Anvil CoderLog in

Use cases

Greenfield project from a specification

A requirement text turns first into a structured specification, then a repository scaffold, and finally into workable epics.

Illustration for the use case “Greenfield project from a specification” (decorative)

Case study in full

Synthetic example — illustrative magnitudes, not a measurement.

Stack profile: typescript-react-vite-spa · Data origin: Synthetic example, not a reference run. · Measurement period: —

A new project starts without existing code, but with a written requirement. From that requirement a structured specification emerges first, then a repository scaffold, and finally a set of workable epics.

Out of scope — not automated: Not automated: the substantive direction of the specification, its approval before the repository is created, and the prioritization of the resulting epics remain human decisions.

Prerequisites

Workflow

  1. A structured, sectioned specification emerges from the requirement text.
  2. Once the specification is approved, the repository scaffold is created.
  3. Epics are derived from the specification and taken into the tracker.
  4. Each epic is broken down into slices and worked, as in the other examples.
  5. The result goes out as pull requests to the new repository, traceable per epic.

Result artifacts

Key figures

Scope

Stories/tickets
5
Slices
22
Files/modules
64

Throughput

Validated slices
20
Retries
7
Human approvals
4

Quality

Checks run
Compile, Unit Tests, Guardrail-Prüfung
Gates passed
6
Gates failed
1
Remaining debt
One epic is waiting for further clarification before its remaining slices are started.

Resources — no pricing

Prompt tokens
600000
Completion tokens
240000
Cache tokens
350000
Model calls
260
Sandbox runtime
100 minutes
Model/stack context
Stack: TypeScript with React, clean architecture (profile typescript-react-clean-architecture); flow Spec Studio → Foundation → Epics.

Log in to see your own run and cost details in the cockpit

Every example on this page was created for illustration. It is not a promise about the duration, quality or cost of a future run of your own — your own figures appear in the cockpit after logging in, within your own tenant context.