Central Casting in practice

How the aimez.ai research program runs on the method.

This page shows the real catalogue and tooling behind the aimez.ai research program, as a worked counterpart to the sanitized demo. The operating layer is Cursor.

The four columns

The program is organized along four columns that the catalogue keeps separate, so any file or action resolves to a clear place.

Instrumentation

The tool layers that produce work, from open-ended reasoning to in-editor drafting.

Actors

The workers inside each lane, named by the lane letter and an instance number, run by the lane's local orchestrator.

Contracts

The system surfaces and written handoffs that say what each lane reads and writes.

ID

The stable index that pairs each lane's local orchestrator, 01 through 016, with its team letter, A through P.

Instrumentation: the drafting layers

OpenAI Chat open-ended planning Cursor operating layer GitHub Copilot inline drafting grounded in repository source

Open-ended planning runs in a general chat so heavy orchestration reasoning stays off the in-editor token budget. Cursor reads and writes the repository, GitHub Copilot drafts inline and every layer is grounded in repository source.

Actors: orchestration and the work lanes

O0 orchestrator Build and validate O1 Authority · O3 Control · O9 Integration · O10 Versioning Research and theory O2 Research · O7 Theory · O8 Topological modeling Product and docs O4 UX · O11 Knowledge · O12 Alignment · O13 Misalignment · O14 R&D Strategy and publication O5 Patenting · O6 Packaging · O16 Publication

00 is the central orchestrator. It routes work to the lanes and reconciles their returns. Each lane has its own local orchestrator, numbered 01 through 016, that records step state for its actors and relays verified summaries up to 00. Orchestration is distributed, so the letter O names the orchestration lane itself and carries no separate worker team.

Contracts: how work moves and remembers

Conversation thread O0 orchestrator intake and routing Work lane task home manifest · README · STEP_LOG · handoffs Checkpoint state recorded a later session reads recent checkpoints and resumes

Every file carries a surface class, so a reader knows from the path alone whether it is an authoritative memo, a derived report, a narrative log or current state. Each lane keeps dated task homes, and a session reads the step log first, so it resumes with the lane history intact.

The full catalogue

The catalogue indexes sixteen work lanes by letter, A through P. The ID is the lane's local orchestrator, 01 through 016, sitting under the central orchestrator 00. The actors that do the work in a lane are named by the lane letter and an instance number, so the research lab lane B runs B1 and B2 and the publication lane P runs P1 and P2.

IDLetterLane
O0zero elementOrchestration and coordinating. Intake, routing, state reconciliation, catalogue hygiene and cross-repo packaging. It does not execute lane work.
O1AAuthority / System / Excavation / Validation / Chain Execution
O2BResearch / Analysis / Experimental Lab
O3CControl System Infra Unification
O4DUX / Documentation
O5EPatenting / Authoring Strategy
O6FPackaging / Boundary Tracking / Presentation Slides
O7GTheoretical Research / Interdisciplinary Collaboration
O8HTopological Modeling / Hatcher
O9IIntegration / Frontend Testing
O10JProduction Versioning / Repo Hygiene
O11KKnowledge Base / ROM Interpretability
O12LProduct Alignment
O13MProduct Misalignment
O14NNew Features R&D
O15OThe orchestration lane itself. 00 is the central orchestrator and 01 through 016 are the per-lane local orchestrators, so the O letter carries no separate worker team.
O16PPublication / Exploratory Findings Manuscript. Presentation, funding and paper-drafting coordination.

How aimez.ai runs on it

The aimez.ai research program is the work this catalogue coordinates, and the clearest single output is the stress-aware routing demo, built and orchestrated through these lanes. The deployment runs along two strands: the site and routing work, and the research grounding and manuscript chain. The cast deployed so far maps onto the lanes like this.

Site, product and routing

Laneaimez.ai work
O0 orchestrationcross-repo orchestration, packaging the program across the site, the manuscripts and the public demos
O4 UX / Documentationaimez site identity, the notes pages and the figures gallery
O5 Patenting / Authoring Strategythe StreetLight validation brief and provider disclosure boundaries
O6 Packagingthe advisor outreach packet and grant proposal revision alignment
O7 Theoretical Researchthe research partner packet, formal theory sharpness and the allocentric flocking bridge
O8 Topological Modelingthe topological framing of the Manhattan graph work
O12 Product Alignmentaimez program identity, holding the program coherent across surfaces

Research grounding and manuscript chain

This strand grew out of the publication chain program, which carried the figure data and the trace reconstruction behind the research. It is where the catalogue has cast multiple actors per lane.

LaneActorsWhat it carried
O1 Authority / ExcavationA1, A2the trace reconstruction lane, excavation lineage and chain execution behind the published figures
O2 Research / LabB1, B2the evidence closure lane and the figure-data chain
O16 PublicationR16, P1, P2the exploratory findings manuscript chain, with R16 the research counterpart and P1, P2 the publication runners

Some lanes are live and some are cast and held. The migration registry in the program repository records the current step for each lane and actor.

Honest scope

The coordinating method was formalized in spring 2026, so it is recent and parts of it are still settling into habit. The discipline reduces context loss across long work and it leans on care more than on automation, so the honest claim is modest: it helps, and it is still being proven.

Walkthrough deck: the slide deck · Full method: WALKTHROUGH.md · Source: github.com/GilRaitses/central-casting