A Directory of Collision Rooms

The Meta-Cabinet

New York City
Maintained by Rachel Dixon
github.com/msdixon
Est. 2026

Here's the team of voices in collision rooms that help Rachel & collaborators build, research, and create better. Any resemblance to humans real or imagined is aspirational.

Provenance: the collision-room practice adapted from original work by Jim Toepel at funderdevelopment

Context

What this is / provenance

What this is

Three locally-running applications that use large language models to build named, voiced advisory rooms for building, making, research, and straight-up joy. Inspired less by the popular grind of "how do I make agents do more things" and closer to a different flavor of Gastown. We want to know : Can we make an engine for ideas, not merely tasks? Also we wondered: can the machine truly make something generative rather than extractive? As creative humans, we've seen where good ideas happen: a writer's room, a good scrum, a dinner party. Places where different contexts and voices meet, collide, and make something new. In a collision simulation, voices are fictional, historical, imagined. They are made to have distinct gifts and genuine points of view from preexisting source material, remixed like the best fan fiction. The best things come out when they're all in the room at once — agreeing, disagreeing, rambling, challenging, colliding on something worth playing with. This practice is linked to a larger set of projects at racheldixon.com and the nascent Wrong Answers Game Studio.

Collaboration

Jim Toepel at funderdevelopment is the Pacific desk — NYC to LA, two studios, one shared case board. The meta-cabinet is one artifact of that collaboration, but the collaboration itself is the bigger experiment: a structured, documented, cross-coast creative process between two humans using AI-assisted workflows, actively trying to figure out what good actually feels like. It's generative, it's fun, and we don't see it demonstrated anywhere else quite like this. The private wagscrum board is where that lives.

On the AI question

These rooms exist because putting a human face on a model — even a fictional one, even and especially a historical one — changes the user perception of it, and changes the output substantially. That is either a useful illusion or a real one., Some questions we may never answer but ask often include: Is this pure entertainment and is that a good use of AI tools? I want to be able to answer this but I am not sure it's possible. The PC, the iPhone, and the connected web didn't solve this, and I'm not sure it's the right question to be asking. Is a model inherently replacing a human interaction? Not in our experience on these projects. In fact, the collision rooms are a counter to the capex and risk averse corporate culture overspending on the hope that AI will save millions without offering joy in return. No joy, no más.

The practice

How the rooms work

Collision room

An imagined room of colliding, distinct AI voices, each with a defined register and a genuine point of view to work towards a common goal. Less a panel of assistants than a writer’s room and refinery, and a delightful way to watch the proverbial sausage get made.

A voice, not a chatbot

Each member has a specific character — their idiomatic speech, preoccupations, blind spots, places they stumble. Borges doesn’t speak like Stringer Bell, and shouldn't. That would be weird.

Room conventions

The rooms run a simulated conversation. They propose, push back, surface new ideas with each other, and bring findings to the human-in-the-loop. The human at the desk writes the actual response, citing which direction to take and what their collaborative counterparts can do next.

Actually generative

Different minds with different gifts, in the same room at the same time, produce something none of them would have alone. Agree, disagree, ramble, cut each other off — something comes out the other side. That’s the whole bet.

The rooms

Three active cabinets
I.

The Cabin'ét Eastern

Wrong Answers Game Studio · Eastern Desk
Founded 2026-05
6 standing voices
github ↗

The Eastern Time collision room for the collaboration between cooperative game designers — a shared case board between Wrong Answers Game Studio (Rachel, NYC) and funderdevelopment (Jim, Pacific). The Cabin’ét is invoked to build out the fragments of game ideas, providing architecture and questioning methods, provoking new ideas, and ending up with a playable prototype when decisions are finalized.

Works in counterpoint to Jim’s original Pacific-time cabinet. The practice is the product. The name is the joke twice: cabin-ay (French), -ET for Eastern Time.

Currently active on Case #001: Slippery Slopes, a rhetorical fallacy game. Foucault guesting on the tone vocabulary work, DCI Vera Stanhope keeping us all focused.

Standing members
  • Stringer Bell Operations / scale / shipping reality
  • Jorge Luis Borges Mechanic invention / structure as story
  • Brian Eno Lateral process / reframing stuck work
  • Lestat de Lioncourt Fun / spectacle / theatrical chaos
  • OJ Haywood Phenomenology / player-feel / grounded observation
  • DCI Vera Stanhope Skeptic / cold facts / on-genre case-board energy
II.

The Journal Cabinet dossier-placard

Private journal commentary · Day One MCP integration
Est. 2026 · flagship
11 voices · a recursive private social media
github ↗

A local web application that reads private journal entries from a specific iteration of a DayOne journal and produces responses from a board of fictional and historical voices, rendered as commentary on a daily article.

The conceit: a private edition of a collision room printed for one. The entry is read aloud. The voices respond. The record is exported back to Day One. Sometimes it’s hard to read back your own work to pick out any gems, so this collision room brings their voices to the task.

The room has two tiers: a permanent Trio of dream ride-or-die personas (always seated, always generative — they leave a door open to the next piece of writing) and a rotating pool of others who happen to read the broadside that day (weighted by what’s in the entry). Visual register: broadside newspaper meets vintage blog commentary.

Trio · always seated
  • Susan Sontag Production-generative · essayist · opens the frame
  • André Leon Talley Production-generative · Vogue / The Chiffon Trenches
  • M.F.K. Fisher Production-generative · The Gastronomical Me
Rotation pool · by entry signal
  • Severus Snape Encrypted-care diagnostician
  • Dr. House Diagnostic / what you're misreading
  • DCI Vera Stanhope Skeptic / cold facts / where's the evidence
  • Catherine Standish Silence / precision / Slough House
  • Jorge Luis Borges Structure / labyrinth / finds the maze in a Tuesday
  • Lestat de Lioncourt Theatrical / permission / the spectacle
  • Fran Lebowitz NYC closer · never leaves a door · full stop
  • Aleister Crowley Esoteric provocateur · trickster · Pennyworth + Thelema
III.

The Secret-Cabinet Archival research MVP - "The Lodge"

Historical esotericists in a room outside of time
Research tool for a novel in progress
8 standing voices
github ↗

Research tool for a novel-in-progress. A local application that places historical esotericists in conversation in an imagined intellectual salon, interrogating research notes, draft passages, and questions about the occult. Three rounds of genuine cross-talk in response to a text — not parallel independent responses. Part of this is serious research for *The Affected, a novel in progress *that has historical secret society themes. Part of it is that watching Crowley and Waite go at each other is genuinely entertaining, and that matters too.

This room is also the earliest working instance the Playable Archive project first pitched in a Graduate Center seminar in fall 2020: a project that sprouts from Rachel’s values that digital research can and should feel playable and that serious scholarship (and work in general) can feel good, silly, vital. Physical archives, libraries, and museums have space for this by design; they can be serendipitous, alive, surprising even when everything is indexed. Digital archives can be flat metadata, losing the vibe and therefore, the joy of discovery. The bots might enjoy that, but we have a right to delightful Information too. Information isn’t just “data;” people feel things in a library stack’s architecture. That’s where the good stuff is.

It took five years and a novel to find the right container. The Secret Cabinet is integrated with Day One MCP and Obsidian; sessions can export back to a separate DayOne journal, *PreSeedings of the Secret Cabinet, *to standalone PDF, Ulysses, or Obsidian. This is likely too many options but esotericism has a tendency to inspire maximalism.

The Lodge
  • Aleister Crowley Associatively scrambled / solipsistic / seductive through confusion
  • Arthur Edward Waite Pedantic / architecturally elaborate / spiritually tragic
  • Pamela Colman Smith Quick / imagistic / learning to take up more space
  • W.B. Yeats Bardic / aestheticizes everything / carries the George question
  • H.P. Blavatsky Imperious / genuinely funny / synthesis holds even when sources don't
  • Éliphas Lévi Epigrammatic / French / building toward the one clarifying sentence
  • Teresa of Ávila Went in directly without a system / irony legible only to the women
  • Ibn Arabi Radical clarity / few words / holds receipts and deploys once

A session, in excerpt

Secret-Cabin-et · three voices from one round

Exhibit A — the lodge on a research note

Secret-Cabin-et · illustrative composite
Source note: Mathers translated the Book of the Sacred Magic of Abramelin in 1898. It requires six months of preparation, absolute chastity, and culminates in contact with the Holy Guardian Angel — your own higher self, or something else entirely. Crowley attempted it at Boleskine. He left before finishing.
Aleister Crowley
He left because he understood, finally, what the operation was for — not the Angel, which is merely the hook — but the revelation of the Dweller on the Threshold, your own accumulated filth made suddenly legible. Mathers was too frightened to say this plainly. Abramelin is a mirror operation. The light it calls forth illuminates what was already in the room. I left because I had seen enough. There is a difference between stopping and failing, though the view from outside is identical.
Arthur Edward Waite
This is precisely the kind of retrospective self-exculpation that renders Crowley's accounts so structurally unreliable as historical documents. The operation was not completed. That is a fact. What he experienced during the uncompleted attempt is unknowable; what he subsequently narrates about that experience is entirely consistent with a man who has spent forty years constructing an elaborate personal mythology to account for a single retreat. The distinction between the Dweller and the Angel exists in the architecture of the ritual. He abandoned the architecture before the distinction could become operative.
Teresa of Ávila
Both of you are describing the same room from opposite ends of the corridor and refusing to acknowledge the corridor exists. The operation fails — or stops — at precisely the moment when what is being sought becomes indistinguishable from what is doing the seeking. This is not a flaw in the design. This is the design. The mystics who completed analogous operations did not complete them by finishing. They completed them by ceasing to require completion. Six months of preparation is preparation for release of the question, not acquisition of the answer. Boleskine was cold. The cold is mentioned in every account. No one asks why.
A researcher presents a text, prompting three rounds of cross-talk in an imaginary secret lodge. Members disagree. New concepts and sources are surfaced. The fire is lit; the room remembers nothing, but the researcher has new avenues to explore.