Skip to main content

Trust

AI Diligence Statement

Where AI helped build Volvelle, what stays human-directed, and what we promise about how it's used.

Last updated2026-06-22Version1.0
On this page

01Why we publish this#

AI tools were used in building Volvelle. That's increasingly true of most software in 2026, but it's worth being explicit about what AI did and didn't do, because users deserve to know what they're using.

Plain English

Short version: Volvelle was created using AI-assisted coding tools. This includes coding agents as well as AI-assisted design. Every feature was planned, directed, and reviewed by a human.

02Where AI was used#

Honest accounting of the places AI assistance shows up in this codebase:

  • Code generation: Claude (by Anthropic) and Codex (by OpenAI) are used as coding assistants. They drive implementations, refactors, tests, and documentation. All coding activity is directed by a human.
  • Design exploration: Claude is used to draft HTML/CSS mockups during the design phase. The final shipped visual language is selected and refined by a human eye.
  • Writing assistance: including portions of this very statement, the Privacy Policy, and the Creator Pledge. Drafted with AI, reviewed and edited with care by a human before publishing.
  • Planning: milestone outlines, architectural docs, User Stories and research summaries are often AI-drafted. Final decisions are human-made.

03What stays human-directed#

  • Product direction. What Volvelle is, what it's for, who it serves, what it will and won't do: those are human decisions.
  • Architecture. Volvelle's framework architecture (the universal-vs-system split, the manifest contracts, the layer boundaries) is human-authored and human-locked.
  • Game content. SRD encoding, character mechanics, manifest authoring: all hand-curated from licensed source material. No AI-generated rules, no AI-generated lore. Where content is taken directly from source books, original licenses apply.
  • Visual identity. The brand, the typographic personality, the aesthetic: these are deliberate human design choices. AI assists with execution, not identity.
  • Every shipped commit. No code lands without human review. AI is a collaborator, not an unattended author.

04AI inside the app (future)#

Volvelle's long-term roadmap includes optional AI features for end-users: character advice, encounter generation, map ideas. When those ship, they will:

  • Be clearly labeled as AI in the UI.
  • Be opt-in, never on by default.
  • Default to bring-your-own-key for transparency about cost and provider.
  • Never use your game content to train external models.
  • Never replace mechanical correctness with hallucinated rules. AI outputs are advisory, the rules engine remains authoritative.
Plain English

Until those features ship, the app you're using contains no in-app AI features. Your characters, rolls, and game state are computed by deterministic code.

05Standing commitments#

  • No AI-generated art in shipped product. The visual style is hand-crafted (code-drawn maps, designed typography) rather than image-model output. If Volvelle ever ships actual graphics, we will use human illustrators.
  • No covert AI features. If AI is doing something in the app, the UI says so.
  • No training on user data. Your characters, maps, sessions, and messages are not training material for anyone.
  • This page stays current. When AI usage changes materially, this page and the changelog are updated.