Make spoken thought portable.
Build a bounded proof of a spoken object that can be captured, replayed, and placed into context without allowing classification to become a precondition for capture.
Capture first. Organize second.
A person should be able to hold a control, speak, release, and immediately possess a durable spoken object. The system may derive a transcript, but the audio remains authoritative.
The proof should demonstrate that an object can exist before its final classification is known. Unresolved objects are allowed; permanent unresolved orphanhood should be visible as an integrity issue rather than silently ignored.
BITE → ANCHOR → ENVELOPE → ROUTE → RENDERER → POLICY
You do not need to implement every layer. Use the architecture to show that the object has identity, can belong somewhere, can move through a system, can be rendered, and can carry rules.
Build the smallest credible proof.
- Hold or press to begin recording; release or clearly stop to complete capture.
- Create a persistent Bite with a visible identity.
- Replay the authoritative audio.
- Generate or simulate a transcript while clearly treating it as derived.
- Assign a Bite to a Family or leave it unresolved/orphaned.
- Represent at least one relationship between Bites or between a Bite and a Family.
- Represent provenance as KNOWN, DECLARED, or INFERRED.
- Represent one lifespan rule such as listen-once, N listens, expiry date, or revoke.
Something we can touch.
- A working URL or locally runnable proof.
- Repository or code package appropriate to the implementation.
- A short architecture note describing major decisions and deferred work.
- A three-minute-or-shorter walkthrough, written or recorded.
- Disclosure of major AI tools, libraries, or generated components used.
The artifact is evidence. Your decisions are evidence too.
- Comprehension — did you understand the actual problem?
- Judgment — what did you deliberately not build?
- Architecture — can another iteration grow from your choices?
- Execution — does something actually work?
- Communication — can another builder understand your decisions?
- Resourcefulness — what happened when the brief was incomplete?
- AI leverage — did tools increase judgment and velocity rather than replace them?
- Integrity — are assumptions, limitations, borrowed components, and unresolved issues disclosed?