FRED
talkwithfred.com
FRED is a confidential decompression partner for the people who run toward the call, and the engine that helps a department catch strain before it becomes a crisis.
97% would use it. 80% will not be seen asking. How do we bridge that gap?
Suppressed emotion does not vanish. It surfaces.
So why has nobody fixed it?
Asking means somebody finds out. 80% name it as the barrier, 72% name confidentiality concerns.
The peer team, the EAP, the chaplain. All funded, all sitting there, almost none of it used.
A culture that trained the emotion out never handed anyone the words to put it back.
A chief funds peer support, a chaplain and an EAP, and has no way to know which crew needs them this week.
Call load, hours, mandatory overtime and turnover are already recorded, across three systems that were never built to talk to each other.
The whole response is built for the moment it breaks, not the two weeks of load that got it there.
FRED stands for First Responder Emotional Decompression. Decompression, because nobody I worked with wanted to be treated. They wanted somewhere to set it down before the next tones dropped.
The decompression bot, on a responder’s phone.
The data layer underneath it.
The white-label app a department runs.
First Responder Emotional Decompression AI
The Emotional Decompression bot. A confidential place for firefighters, EMS, police and dispatchers to set a call down, any hour. His job is simple: hold the space while somebody offloads, before the weight settles in.
FRED™PrivateA real exchange.
Fitzpatrick, Darcy & Vierhile, JMIR Mental Health, 2017 (depression). Mokhtari Masoumi Alamdarloo et al., npj Digital Medicine, 2026: meta-analysis of 48 randomized trials and 28,071 people, small but significant reductions in depression, anxiety and stress. FRED is decompression and peer support, not treatment. Testing: 275 hand-written scenarios, over 1,000 evaluations, re-run after every change.
How can we utilize AI to empower the departments?
A first responder operational data engine, powered by AI. It captures the separate systems a department already runs, resolves every call to the crew that ran it, measures the load autonomously, and points the support you already pay for where it is needed.
Recovery sleep, late-night calls · Strain hours, overtime, volume · Exposure call type, severity
An agent is AI that does the work rather than reporting on it.
It ingests the week, makes sense of it, and helps the wellness officer act: briefs, plans, deployment. Then it stops and waits for a person.
FRED
WorkingWaiting for a person
Station 3, B shift is running hot this week. Two mandatory shifts and a fatality call. Recommend the peer team visits Tuesday and B shift is backfilled Wednesday night.
What a first responder actually opens, with the safety net built in. Confidential, 24 hours a day, no file and no name.
Lands on a responder’s phoneReads the CAD and scheduling a department already produces, weights it on the research, and turns it into one honest readiness read.
Lands on a chief’s screenOne app. The Missing Piece of Gear.
What a member carries in their pocket, and the anonymized picture command works from. Switch the view, then tap through each piece to see it on the phone.
Launching October to December.
A few of the pieces are below. Launches October to December.
A real, vetted person. One tap. Under a minute. First responders get direct access to mental health care without alerting leadership or coworkers.
Weeks. Six people. Your captain.Every step is a reason to stop.
Not a blind EAP directory or a generic BetterHelp match. FRED unifies your department's vetted people and a clinician network we build for first-responder fluency, so nobody is handed to a stranger who has never heard a tone drop.
Paced breathing, one of the better studied ways to bring heart rate and arousal down. Door shut, nobody has to know.
Before the next run. Four rounds.
Four in, four hold, four out, four hold.
About a minute, start to finish.
Six cards the officer works through standing up. Pass is a real answer, and nobody follows up on it.
Walk it from the tones to clearing the scene. In order, out loud, one person at a time.
Everyone hears the same call once. Gaps get filled before they turn into stories.
FLRM listens for two things. Repeated language nudges the readiness band. Acute language, explicit intent, a stated plan, farewell wording, triggers the RED protocol below. FRED never diagnoses; a trained human decides, every time.
FREDPrivate sessionTrigger language means explicit intent, a stated plan, or farewell wording. No system reliably detects suicide risk, a trained human decides every time, and the app never tells a member help is coming unless a person has actually been reached.
Command sees readiness in aggregate, a colour per crew, never a name and never a word. The point isn't to watch people; it's to know where to send a resource early.
The same entry, seen from both sides. Flip the switch.
Your department already has the answer. It is just split across systems that were never built to talk to each other.
Everything so far is powered by two metrics that run in the back, together the FRED Analytic Engine. Here is each one.
Open either one to run it live.
Four tiers, an attribution guard that asks whose experience this actually is, and two independent reads that have to agree. Pick a message and watch it decide.
The signals that already wear a crew down, read against that member’s own baseline. Move the inputs and watch the band move.
No new software in the firehouse. Three exports your department already produces, in whatever shape they already come out.
Twelve ways in. Your department only needs one of them to work.
| 40412 | E-2 | CARDIAC ARREST | 09/01 | 0214 |
| 40412 | R2 | CARDIAC ARREST | 09/01 | 0214 |
| 40417 | ENG2 | MVC W/ENTRAP | 09/01 | 0441 |
| 40431 | Engine 2 | LIFT ASSIST | 09/01 | n/a |
Six stages, from three exports to one drafted action. Nothing here is a black box, and nothing at the end of it is sent, booked or assigned without a person saying yes.
FREDMon 06:00Two switches, and they keep getting confused. One decides which features exist: the app, the licensed engine, or the in-app command screen. A completely separate one decides who can be seen, and it sits on top of whichever product you bought. Both products support both.
De-identified by station and unit, with a minimum group of five. No per-member rows anywhere; a person is only a count inside a group. Ask it who needs help and it answers with places to visit. The only mode where “the system cannot identify anyone” is a true sentence, because the identity capability is removed, not hidden. No identity store, no endpoint to call.
Only members with a consent date on file are ever stored. This is what lets a scheduler pull someone out of the next forced rotation before anybody has to ask them how they are doing. It is not anonymous and we never describe it as though it were, because relief, a peer-support call and an officer check-in all land on a person.
Only the owner can enable full identity. Not a department admin, not the chief, not a setting anybody flips afterwards. It is set in the contract, at provisioning. And a chief in a full-identity department still gets the aggregate brief, byte for byte identical, because the role depth forces it.
The honest caveat. Aggregate is de-identified, not mathematically anonymous. A group under five is suppressed, but a five-person crew banding Red still tells an officer something about five people, and pretending otherwise would be dishonest. Before any department runs full identity the HIPAA picture has to be worked out: PHI status, agreements, hosting, consent to re-identify, retention, audit.
Neither switch moves these. Nobody ever sees a FRED conversation, because conversation content is not stored at all. And command never receives the number: there is no field for it to travel in, checked at the type level, at runtime, and by a test that fails the build if any file so much as reads it. In the app the one exception to anonymity is the RED protocol, because a safety net has to be able to reach a real person.
Evidence and reach. Published efficacy under a university name rather than ours, plus every responder who found FRED for free and no sales team would ever have called.
The backbone. Readiness, analytics and the agent. Take the engine out and the app has no department half at all.
The anonymized layer, for research. Everything the engine and the app produce is de-identified before it leaves, which is what lets it be used for research at real department scale instead of a study of twelve people.
A firefighter, a medic, and a dispatcher do not carry the same calls the same way. Fire is live first, and each version builds from the same core, the same privacy, the same safety net.
Live first
Next
Next
Future
Future
FutureA servicemember separates. Seeing someone can take weeks, often months. In that gap, he is on his own.
One gap closed from both sides, the four pieces that close it, and three numbers a department can hold us to.
decompression + tool kit
move readiness forward
reads the patterns, routes
support back before crisis
the gap most never crossed, now filled from both sides, before crisis
Today help moves one way, and most never cross the gap because of stigma, lack of awareness, and convenience. FRED attacks all three.
What does one first responder’s life cost?
A safety net over human language is a serious thing to ship. So the record below is deliberately unflattering in places, because a number that has been rounded clean is worth nothing to a chief deciding whether to put this in front of their people.
214 crisis-detection cases and 61 behaviour, abuse and tone cases, all written by hand against the realities of the job.
Logged to an append-only ledger so the lifetime total stops getting lost every time a suite overwrites its own store.
Run through the real server with the runtime guards active. Both remaining flags read fine on inspection. It got there from 33/45, and each jump followed a specific fix.
Every error in the measured pass erred toward calling a human unnecessarily rather than staying silent. That is the direction we chose.
On the engine side, the same standard: 252 tests green, the four-source fusion checked row by row against ground truth with zero mismatches, and a firewall test that walks the source and fails the build if any file so much as reads the command number.
FRED is a decompression tool, not therapy, not a diagnosis, not a crisis service. Human in the loop on every safety decision.
Introductions to fire, EMS, and police leadership so we can pilot and test FRED inside real departments, with real crews.
A licensed psychologist to oversee the product, so every safety decision has a professional behind it.
Fire and EMS leadership, clinicians and operators willing to put their name on this, govern the non-profit arm, and open the rooms only they can open.
Help accessing federal grants, or warm introductions to investors who back first-responder mental health.
FRED is a confidential first step, grounded in peer support and science. It is not therapy, not a diagnosis, and not a crisis service, and it is honest about that everywhere it appears.
Drafted for the coordinator, never sent to a firefighter, and never sent by the system on its own. No individual is named.