CASE STUDY · ARMR

The question they dread, answered fifty times before it is ever asked.

CLIENT
Dark Horse Political
WHAT WE BUILT
The product, the AI, the voice pipeline, the brand and the site
DISCIPLINES
Software · AI · Voice · Brand · Web
STATUS
Live at getarmr.ai

THE PROBLEM

One question. One camera. One eight-second clip.

A year of work, forty thousand doors and the trust of everyone who believed in a candidate can be gone before they finish a sentence. The clip doesn’t care that they were tired or that the question was unfair.

Top preparation firms charge $25,000 to $75,000 a cycle and can’t be in the room when it matters. Public-speaking apps grade your “ums” and have never heard of your opponent. So most candidates walk into the most important moment of their campaign having never once rehearsed it under fire.

WHAT WE BUILT

A simulator that knows the race better than the moderator does.

ARMR is invite-only adversarial training. Before the first drill it researches the candidate’s race: the district, the opponent’s record, the lines of attack already in play. Then it puts the candidate on the spot, out loud and on the clock, in a room called The FORGE.

  1. 01
    It builds the real opponentAn AI research pass assembles an intelligence profile of the race and turns the opponent’s actual record into the questions they will actually face.
  2. 02
    The candidate answers out loudSpeech in, voice out. Questions are delivered by a cast of questioners, answers are spoken and transcribed, and the clock never stops for a second draft.
  3. 03
    Traps, follow-ups, pressureSessions escalate from warm-up to hostile. Traps fire when the candidate takes a premise they should have refused, and follow-ups arrive when an answer leaves a door open.
  4. 04
    Every answer is scored and coachedSix dimensions on every answer (composure, message discipline, pivot execution, logical consistency, attack deflection, affirmative framing), with a model answer and a coach’s note.
  5. 05
    A human signs offA senior consultant reviews the record and signs a readiness readout. The machine trains; a person decides when the candidate is ready.
An ARMR session post-mortem: a timestamped transcript of moderator questions and the candidate's spoken answers beside a six-dimension scoring panel and a coach's note.
The post-mortem. Every question, every answer, the clock, the six scores and the coach’s note, the moment the session ends.
The candidate dashboard: composite readiness 7.4, 23 sessions, an 11-day streak, a rising readiness chart, a six-dimension radar and a consultant's signed readiness readout.
Readiness. Progress over thirty days, the six dimensions, and a consultant’s signed verdict.
The drills screen: full session, target my weakness, self-command, frame control, read and recognise, offensive strikes.
Drills. Targeted sessions, including one that attacks the candidate’s weakest dimension on purpose.
The Strategy Hub: a library of frameworks such as vulnerability mapping, trigger recognition and pivot mechanics.
The Strategy Hub. The doctrine behind the scoring, written down so the candidate can study it between sessions.

THE ROOM

Twelve questioners: rivals, moderators, local and national press, constituents, and the incumbent herself. Questions are read aloud and answered aloud. The candidate is never typing into a text box.

The ARMR logo: a single signal line that drops and climbs, above the ARMR wordmark and the words Adversarial Response and Media Readiness.

The identity we built for it: one line under pressure, carried through the product, the site and every drill.

337AI-written questions audited line by line
0invented places found in that audit
6dimensions scored on every single answer
4,399automated tests guarding the product

Screens are from the live product, shown with a fictional candidate and race.