usertold 3.0.0

A simpler model from research to product decisions

UserTold now presents one consistent mental model across the CLI and the service:

• Interviews are the source record of what a participant said and did.
• Evidence is the source-linked analysis extracted from those Interviews and kept reviewable against the original record.
• Findings synthesize related Evidence into a problem worth evaluating. A Finding can be reviewed and prioritized without implying that a solution has been chosen or delivery has been committed.

This separation also makes the lifecycle clearer. Research review establishes
whether a Finding is supported and ready for product consideration; delivery
tracks what happens only after it is intentionally sent to GitHub or Linear.
Research maturity and product execution are related, but they are not the same
state.

To make that model real rather than a vocabulary layer, 3.0.0 aligns the public
interface with the canonical internal implementation. usertold findings now
uses the canonical Finding API and DTOs directly, help follows the complete
Study → Interview → Evidence → Finding journey, and tracker selection defaults
to the Project configuration in the dashboard while allowing an explicit
provider override.

Migration from 2.x

• usertold work is now usertold findings; there is no work alias.
• The Evidence list filter --work is now --finding.
• Finding JSON uses the canonical Finding response model rather than task-shaped presentation keys.
• Opaque references remain pass-through values, and existing 2.x installations continue to work through the service's retained compatibility API.

The packaged JSON help, shell-facing command metadata, documentation, and HTTP
User-Agent now describe and report this same 3.0.0 interface consistently.

This is a major release because the command-group, filter, and JSON model changes
are intentionally breaking for scripts written against 2.x.