A mind, walked from the inside. Five districts stand around a lit beacon, each holding one of the notes below out in the open — walk up to one, or read straight from the list. The ground is a small planet, so it has no edge: keep walking in any direction and you come back to where you set off. One of the five, The Room, has a door you can step through: inside, a single note drifts overhead as light you can reach on foot.
4Color.AI starts from a mathematical image: a complex map can be separated with only four colors if the boundaries are understood.
That is also a useful way to think about AI systems. The hard part is not adding more color. The hard part is finding the right boundaries.
MirrorX is a persistent agent-world research project. It studies whether agents can live inside real software environments, carry memory across runs, and expose the gap between what they believe and what the world currently says.
MirrorX has reached a live-testable V2 spine. The first proving ground is KickStart staging world-1, a sandboxed SaaS environment with synthetic accounts and isolated org data.
The current spine can run controlled agent jobs against a real host app. The first live-testable job is coach-owned drill and assignment work inside KickStart. The agent acts through an app persona, mutates sandbox state, writes persistent handoff artifacts, and leaves audit evidence for later runs.
Latest proof passed:
The status is not "finished platform." The precise claim is narrower: MirrorX now has a first real app sandbox loop.
The original MirrorX brief described a future observatory for agent societies. The new milestone is narrower but more important: a real app loop exists.
--use-latest-loop-state to select the newest ready loop stateA scripted demo can prove a path works once. MirrorX is trying to prove whether persistent agents drift, recover, remember, misremember, and act inside mutable app worlds.
The research value is not automation by itself. The research value is making agent/world mismatch inspectable.
MirrorX separates the agent's durable body from the model backend currently driving it:
That separation matters because the agent may have multiple jobs and roles over time. "Coach" is one job in one host app, not the whole agent identity. The long-term system should let model backends change without losing the agent's body, world state, or action history.
MirrorX is not a complete framework yet.
KickStart is the first proving ground, not the final product.
The Four Color Theorem begins with a simple question: how many colors are needed so neighboring regions never match?
That question becomes a doorway into representation, constraints, search, and the habit of turning messy reality into a structure a machine can reason about.
An agent is not interesting because it produces text. It becomes interesting when it can remember, choose, revise, and act inside a shared environment.
MirrorX uses that idea inside a real app sandbox. In the current V2 spine, an agent can act through a KickStart coach persona, create a coach-owned assignment, persist loop state, and leave evidence that a parent persona can see the result.
MirrorX separates two ideas that are often mixed together:
The body is what lets an agent persist inside a world. The spirit is the model currently driving decisions. A useful persistent agent should be able to change model backends without losing app identity, world attachments, memory, or audit history.
In a toy simulation, state can be whatever the experiment says it is. In a SaaS sandbox, the world can disagree.
An agent may believe it created an assignment. The app may have rejected it, hidden it behind permissions, or made it visible only to one role. MirrorX turns those mismatches into inspectable evidence:
That is why the KickStart world-1 loop matters. It is not the final multi-agent society. It is the first real test of whether agent memory, app state, and user-visible evidence can stay reconciled.
Logical thinking looks deterministic from the outside: premises, rules, conclusions.
But useful reasoning often depends on search, sampling, surprise, and controlled variation. This trail explores whether logic is less like a railroad track and more like randomness under discipline.