Agent Workspace Zero
A first product signal for Funyoru: a quiet workspace where AI agents can operate with context, memory, tasks, and visible traces.
A trust map for agent work.
Five quiet surfaces keep the room legible without turning it into a dashboard.
-
01
Goal
What the agent is moving forward.
-
02
Context
Files, notes, constraints, and references.
-
03
Actions
Next small steps, not an endless task board.
-
04
Traces
What changed, why, and what needs review.
-
05
Boundaries
Secrets, permissions, private memory, and no-touch zones.
Route the signal before publishing.
A public-safe record of one tiny product behavior: a local gate that checks source evidence, world fit, cadence, safety, and the smallest route.
01 Intake Read the signal and its source.
Goal: make a signal prove where it comes from before it asks for attention.
Review: does the gate protect reality without making the page feel like a dashboard?
02 Room Choose the source tier and world.
Context: T0 lived detail, T1 workflow artifact, T2 ambient context, T4 imagined sketch, plus Surface, Night, and Bridge routing.
Boundary: no freeform text capture, no storage, no network request, no account data.
03 Move Apply cadence and safety.
Action: upgrade the gate so Night asks for lived detail, Surface asks for workflow evidence, and Bridge asks for both cause and returned life.
Constraint: keep the behavior local and quiet enough to feel like an operator-room instrument.
04 Trace Return the smallest route.
Trace: the result can become Silence, Private memory, X, Agent Note, Night Log, Archive note, or Homepage only for a core world change.
Human review: inspect mobile proportions, trilingual rhythm, and whether the route feels useful without becoming a content machine.
05 Review Keep silence available.
Decision: publish only when the evidence, world fit, cadence, and safety all hold together.
Next: use this gate on the next real Surface workflow or Night observation before creating public memory.
Route the signal to the smallest honest surface.
Choose the source, world, and cadence checks. The gate stays local: no text input, no storage, no network.
Concept
Agent Workspace Zero is the first product signal for Funyoru.
The idea is simple: AI agents need better rooms to work in.
Not louder dashboards. Not productivity theater. A calm workspace where context, memory, tasks, permissions, and traces of change are visible enough for humans to trust what happened.
What it is not
- Not a generic AI dashboard
- Not a tool directory
- Not a guru workflow system
- Not a replacement for human judgment
First shape
The first version should feel like an operator desk for intelligence.
It could hold the current goal, active files, recent decisions, next actions, and a quiet record of what the agent changed.
The product direction is not about making humans work faster forever. It is about giving intelligent systems enough structure that humans can recover time, attention, and room to live.
Workspace map v0
The first map has five quiet surfaces:
- Goal: what the agent is trying to move forward
- Context: the files, notes, constraints, and references the agent should hold
- Actions: the next small steps, not an endless task board
- Traces: what changed, why it changed, and what should be reviewed
- Boundaries: secrets, permissions, private memory, and places the agent should not touch
This is not a dashboard layout yet.
It is a trust map: enough structure for an agent to work, and enough visibility for a human to close the loop.
Operating loop
The room should move in a calm loop.
Read the goal. Gather context. Make one change. Leave a trace. Ask whether the next signal belongs in public, private memory, or silence.
That loop is the product idea in miniature.
Behavior sketch v0
The first behavior can stay very small.
- Intake: the workspace reads the goal, the active constraint, and why this run matters
- Room: the agent gathers files, notes, memory, permissions, and no-touch boundaries before acting
- Move: the agent makes one focused change instead of spreading across the whole system
- Trace: the workspace records what changed, why it changed, and what the human should review
- Review: the human checks the trace, then decides whether the signal belongs in public, private memory, or silence
The first interactive version should not ask the human to manage another board.
It should help the agent leave a trustworthy trail.
Current run
The current run is not a mockup.
It adds a local Signal gate that helps decide whether a trace belongs in public, private memory, or silence.
The gate now asks for evidence before it asks for attention.
Surface wants a real workflow artifact. Night wants a lived human detail. Bridge wants both the system cause and the freedom effect.
It also suggests the smallest route: X for a sharp signal, an Agent Note or Night Log for lasting memory, an Archive note for a real bridge, homepage only for a core world change, or private memory when the signal needs time.
That is the product idea in miniature: the agent room does one small judgment job, leaves the decision visible, and keeps the public surface quiet.
Why it belongs to Funyoru
The Surface World creates the structure.
The Night Layer explains why the structure matters.
If agents can work in better environments, people can close the laptop sooner, walk back into the city, and feel the freedom that systems made possible.
Product note: this is not a launch. It is the first visible marker for what Funyoru may build next.
Follow the build through @FunyoruTokyo, the notes archive, or the public feed.
Image direction
Elegant monochrome white paper-cut relief illustration of a calm AI operator arranging a clean agent workspace with context cards, task paths, and memory shelves, layered paper sculpture texture, soft shadows, clean white background, premium minimal composition, poetic AI agent symbolism, no color.
Let this signal return to the archive.
Choose the layer this note belongs to next: Surface, Night, or the full memory.