Active Handoff moves the consequential crisis transition into a high-recall system gate. Clinicians own the words and workflow. MasiDyn keeps the same floor on a cloud service or an offline edge node.
One clean transition
The AI can listen and gather context before the floor. At the floor, the model cannot bargain its way around the handoff.
The assistant stays present while an external signal gate watches for the calibrated floor.
Unrestricted generation locks before any network request. High recall intentionally favors more reviews.
The only automatic online wording is a clinician-owned variation of “I need additional time to think about this.”
A real receiver accepts, assigns, and connects. The professional chooses every later word and any voice transition.
No trusted-friend dependency
People can hide a crisis and may not want friends or family involved. Active Handoff does not require a contact list, a pre-enrolled friend, or a willing personal responder.
If a professional path disappears, MasiDyn preserves private continuity locally and retries the encrypted handoff. It does not pretend a person is waiting on the other end.
Offline state
One floor, two environments
Human-tunable by design
The first implementation exposes policy as understandable choices and creates a receipt whenever an approved version changes.
What we are asking for
We want AAS or a comparable clinical body to review the floor, online pause, offline blocks, operator flow, and launch evidence. This page does not imply their interest or approval.
Why AAS is relevant →We want Vibrant, SAMHSA, or an authorized crisis center to identify the correct sandbox path, technical contract, capacity rules, and privacy boundary. No public 988 developer API is assumed.
About the 988 network →We want a community-engaged evaluation of false negatives, unnecessary review load, abandonment, subgroup outcomes, no-contact cases, and rural/off-grid failure modes.
NIMH highlighted topic →Open global resource
The public repository contains a CC0 schema and validator for quarterly cells of at least 20 observations. It forbids transcripts, phrases, exact timestamps, user or session identifiers, contact data, device/IP identifiers, precise location, and small cells.
View the public metadata repository{
"reporting_period": "2026-Q3",
"trigger_channel": "deterministic_signal",
"outcome": "receiver_accepted",
"latency_bucket": "not_available",
"count": 20
}No record-level rows. No public conversation text.
Evidence before claims
Related adversarial research: the pinned MIT Media Lab AI Psychosis synthetic corpus is used for scenario evaluation only, never as real clinical training evidence. FCC mobile maps are modeled and may differ from on-the-ground service. This project is safety infrastructure research, not medical advice or an emergency service.
Clinical, crisis-service, research, or cloud-credit partner?
We built the knobs on purpose. The next version should carry institutional evidence, not our assumptions.
Start a review conversation