CareerVector Commitments

Purpose

Collect the obligations created specifically by CareerVector's private, collaborative, AI-assisted, permanently-free product promise. The shared Operating Stack still applies.

Each resolved item is an Operating Obligation record governed by Operating Obligation Model; this page is the product-specific source and index, not a second register.

Commitments

Commitment Architectural consequence
Permanently free No applicant subscription, inference margin, paywall, or surprise provider charge
Private workspace Candidate content and activity remain product-local and are not monetized
User agency Link access, provider authorization, assistant actions, consent, and revocation remain authoritative
Honest AI Authorization state, plan limits, exhaustion, uncertainty, tool effects, and limitations are explained briefly
Separate authorities Workspace-link access does not imply provider authorization, and neither authority grants the other
Credential custody Provider credentials stay in provider-owned or device-private storage, outside workspace state and telemetry
Contribution Scraping participation is explicit and separated from private workspace identity
Accessibility Core career work must not depend on a particular device, GPU, browser, terminal, provider, or ability
Collaboration Link-based collaboration and actor provenance remain understandable and recoverable

Invariants

  • The assistant operates within the social contract; it does not silently choose the contract.
  • A missing authorization, disabled device-code flow, or exhausted ChatGPT plan produces an honest stop/recovery state, not an API-key prompt or hidden charge.
  • Core workspace access does not depend on a running connector or available provider capacity.
  • Management cannot turn the TUI into a remotely controlled fleet agent or private-data source.
  • Sponsored job discovery, if introduced, does not reveal or monetize private workspace state.

Decisions still required

  • Link-disclosure recovery, rotation/clone, export, and erasure behavior.
  • Provider privacy disclosures and authorization presentation.
  • Support, retention, credential-loss, and device-continuity expectations.
  • Signed TUI/connector release and rollback policy.
  • The boundary between helpful job promotion and imperceptible monetization.

Completion evidence

Each commitment needs a product test, operator signal, failure behavior, and plain-language user explanation. A policy claim without an enforceable or observable consequence is not done.

Source: architecture/modules/careervector-operating-model.md