Working demo · intelligence density harness
Build a verifier-gated Bonsai pack
Describe the workflow, watch the pack draft beside you, approve risky writes, and keep claims inside the proof envelope before anything leaves the boundary.
Wireframes 1A · 1B · 1C
Workflow interview to draft pack
I run care-navigation intake. We need to capture synthetic referral notes, remove identifiers, and send only clean triage themes to the team.
I hear three signals: member identifiers, contact details, and family details. I will draft recognizers, a clean record schema, default-deny egress, synthetic eval rows, and safe claims.
The value story is internal adoption. Show leadership that the sensitive note stays local while the clean record reaches Slack.
Pack draft ready in scratch space. It needs human approval before writing files or preparing an upstream benchmark report.
Draft assembles alongside the interview
The agent converts workflow pain into recognizers, policy, schema, eval rows, safe claims, and a benchmark report.
Evidence before claims
Verifier and benchmark state
Synthetic data only
3 eval rows, 0 real notes, 0 raw examples in publish copy.
Residual identifier scan
0 residual identifiers in clean Slack payload.
Default-deny egress
Slack sink requires verifier.residual_identifiers == 0.
Claim critic
Unsafe terms are blocked until copy says scrubbed or redacted.
Prism benchmark export
Report includes hardware, model family, eval size, limits, and gate result.