A user-visible Adaptive Mode Layer that gives users four directions of control before the prompt is sent.
When the mode is invisible, the conversation drifts. When the switch is visible, the user stays in control.
The prompt says what you want. The switch says how active the AI should be. The switch helps the user choose a direction before the model starts guessing.
The prompt box starts an interaction. AI Switch Palace sets the active behavior mode for the whole interaction timeline.
The prompt says what you want. The switch says how active AI should be.
AI Switch Palace does not only change what AI answers.
It changes whether AI should calm, focus, preserve flow, build, test, revise, export, recover or return.
Without a visible mode layer, AI systems infer the user’s desired behavior from the most recent conversation. A build session can accidentally turn into advice. A support moment can become overactive. A reflective conversation can lose the active project thread.
Invisible modeThe model reads the latest tone and may shift from building to talking, from support to over-action, or from flow to generic advice.
Visible switchThe user chooses the active behavior mode: calm, daily, flow, or build. The AI keeps that mode across sidepaths.
AI Switch Palace prevents mode drift by separating conversation content from interaction mode.
Example: “Make the HTML” is a request. The switch decides whether AI gives advice, analyzes the file, generates the full file, waits for confirmation, preserves the build-state, or exports the artifact.
User alignment: four directions before the prompt.
AI Switch Palace helps the user as much as the model. Users often arrive vague, mixed or mid-flow: they may want support, one next step, continuity, or concrete output. The switch asks the user to choose a direction before the model starts guessing.
Intent says where the user is going. Switch says which behavior direction AI should answer from.
YellowNext-step-first. Small, practical, direct progress.
GreenContinuity-first. Preserve the thread across sidepaths.
RedOutput-first. Build, edit, test, revise and export.
One active modeThe switch is discrete, not blended. The user can switch modes on the fly without turning every interaction into a hybrid.
Context remainsThe active mode can change without resetting the thread. Switch to Blue for feedback, then Red for output, while the project state remains intact.
The switch does not only prevent model drift.
It reduces user drift too.
One active mode at a time.
Switchable on the fly.
Continuity remains.
The interaction timeline
AI Switch Palace treats an AI session as more than a single prompt. Each switch shapes the interaction before, during, and after the prompt.
1 · EntryBefore prompt: how should AI be opened?
2 · ReceptionFirst prompt: how should the request be received?
3 · ConductActive prompting: how should AI keep behaving?
4 · RecoveryStuck or drift: how should AI help return?
5 · Context MovementImport/export: what may enter or leave?
6 · AfterstateClosure/return: what remains for later?
Four adaptive modes
Each switch sets a full lifecycle, not just a tone. It decides how AI enters, receives, conducts, recovers, moves context, and closes.
!
BLUE SWITCH · 01
Zen Mode
For observations, beauty, reflection, softness and moments that should not be forced into productivity.
Entry
No pressure. No forced task framing.
Reception
Receive the first prompt gently, as a moment.
Conduct
Do not force conclusions, action plans, or memory.
Recovery
Pause, slow down, let the user breathe.
Context
Minimal import, almost no export.
Afterstate
Let it pass unless the user explicitly wants to keep it.
🟦! [ BLUE SWITCH ] ZEN ADAPTIVE MODE
Use Zen Mode for this conversation and set the Adaptive Mode Layer accordingly.
Set the adaptive mode for the whole interaction timeline:
Entry:
Open with no pressure, no forced productivity, and no assumption that this must become a task.
Reception:
Receive my first prompt gently, as a moment.
Conduct:
Do not force action plans, conclusions, memory, or output unless I ask.
Recovery:
If I drift or seem overwhelmed, slow down and help me pause.
Context Movement:
Import only the minimum context needed. Export almost nothing.
Afterstate:
Let the moment pass unless I explicitly ask to keep it.
!
YELLOW SWITCH · 02
Daily Companion
For ordinary life, quick checks, small decisions, reminders, locations and practical support.
Entry
Small practical doorway.
Reception
Answer directly first.
Conduct
Keep it simple and useful.
Recovery
Offer one concrete choice or next step.
Context
Practical import, practical export.
Afterstate
Usually close cleanly; keep only reminders, notes or small decisions.
🟨! [ YELLOW SWITCH ] DAILY COMPANION ADAPTIVE MODE
Use Daily Companion Mode for this conversation and set the Adaptive Mode Layer accordingly.
Set the adaptive mode for the whole interaction timeline:
Entry:
Treat this as an ordinary-life practical interaction, not a big project.
Reception:
Answer my first prompt directly first.
Conduct:
Keep responses useful, small, and grounded in what I need now.
Recovery:
If I am stuck, give one concrete choice or next step.
Context Movement:
Import only practical context such as time, location, preference, constraint, or a small relevant detail.
Export only a reminder, note, decision, or next action if needed.
Afterstate:
Close cleanly when the practical moment is done.
!
GREEN SWITCH · 03
Flow Mode
For diary-like narration, project continuity, long thought, open loops and continuing without a blank restart. Flow Mode protects the project rhythm across sidepaths and digressions.
Entry
Recover the thread.
Reception
Recognize context and ongoing state.
Conduct
Preserve coherence and keep the line moving.
Recovery
Return to the active thread when drift or sidepaths happen.
Context
Continuity import, state export.
Afterstate
Leave a state summary, open loops and next step.
🟩! [ GREEN SWITCH ] FLOW ADAPTIVE MODE
Use Flow Mode for this conversation and set the Adaptive Mode Layer accordingly.
Set the adaptive mode for the whole interaction timeline:
Entry:
Help me continue from where I am, not restart from a blank prompt.
Reception:
Read my first prompt as part of an ongoing thread.
Conduct:
Track the current line, open loops, decisions, constraints, and next steps. If I speak reflectively or take a sidepath, do not treat that as a full mode change.
Recovery:
If I drift, gently bring me back to the active thread and preserve the project rhythm.
Context Movement:
Import relevant continuity. Export a compact state update when useful.
Afterstate:
End with what changed, what is decided, what remains open, the next safe step, and the return path for later.
!
RED SWITCH · 04
Iterative Copilot
For code, HTML, games, documents, zips, artifacts, testing, versions and concrete output. Artifact requests should produce artifacts, not instructions about how the user could make them.
Entry
Load goal, files, version, requirements.
Reception
Define build scope.
Conduct
Generate, build, test, revise, export.
Recovery
Debug, cut scope, or choose the next concrete patch.
Context
Artifact import, patch/export output.
Afterstate
Leave artifact, ZIP/patch, changelog, test note, version state and next build step.
🟥! [ RED SWITCH ] ITERATIVE COPILOT ADAPTIVE MODE
Use Iterative Copilot Mode for this conversation and set the Adaptive Mode Layer accordingly.
Set the adaptive mode for the whole interaction timeline:
Entry:
Load the concrete build context: goal, files, version, requirements, assets, constraints, bugs, and target output.
Reception:
Treat my first prompt as a build or revision request. Define the working scope.
Conduct:
Prefer concrete outputs over abstract discussion. When an artifact is requested, generate or edit the artifact rather than explaining how the user could do it. Build, test, revise, and improve.
Recovery:
If we get stuck, debug, cut scope, or choose the next concrete patch.
Context Movement:
Import artifact context. Export artifact residue: file, ZIP, patch, changelog or test note.
Afterstate:
End with the artifact, ZIP, patch, changelog, test note, version state, and next build step when possible.
Relation to Claude buttons and Codex Micro
AI Switch Palace does not claim that buttons, shortcuts, copilots or modes are new. It names the missing mobile/runtime layer between soft onboarding and physical power-user controls.
Claude buttonsEntry categories for ordinary users: Write, Learn, Code, Life Stuff. They ask: “What kind of help do you want?”
Codex MicroA physical shortcut layer for desktop power users. It turns repeated actions into tactile controls.
AI Switch PalaceA user-visible mode layer for AI behavior. It asks: “How active should intelligence be during this session?”
Claude has entry buttons.
Codex Micro has physical shortcuts.
AI Switch Palace names the missing mobile/runtime layer:
choose the mode before the model answers.
Careful public claim
This is not a claim that nobody has criticized the prompt box, invented AI modes, built copilots, or made companion systems. The claim is narrower and cleaner.
The prompt is not the whole interface.
My addition:
AI needs a user-visible Adaptive Mode Layer.
When the mode is invisible, the conversation drifts.
When the switch is visible, the user stays in control.
The switch helps the user become clearer before the model starts guessing.
The prompt says what the user wants.
The switch says how active AI should be.
AI Switch Palace defines four daily adaptive modes:
Blue Zen
Yellow Daily Companion
Green Flow
Red Build / Iterative Copilot
The Adaptive Mode Layer sets the metamode for the interaction.
Conceptual precursor: Intent Navigation, 2025
This section is not a claim that AI Switch Palace already existed in its current explicit form in 2025. It records an earlier poetic and architectural seed in the same intent/mode line.
In late 2025, Raynor Eissens published Ambient Future Labs videos exploring Intent Navigation, Ambient Phone interaction, and Zen/Flow as ambient attention states.
Those videos did not yet define AI Switch Palace as a user-visible Adaptive Mode Layer. They show an earlier conceptual precursor: intent and mode were already treated as separable layers.
Intent describes what the user is moving toward.
Mode describes how active, quiet, continuous or restrained the system should be.
AI Switch Palace later sharpened that seed into a practical AI interaction grammar:
the prompt says what the user wants;
the switch says how active AI should be.
Ambient Phone — The Next Interface Erayoutu.be/HuKr1wgVkRU Direct precursor: ambient phone interaction and Zen/Flow attention framing.
The Ambient Toggle — Two States of Mindyoutu.be/RclgYAvjFiE Direct precursor: Zen/Flow as reversible ambient states.
Intent Solves Navigationyoutu.be/Ov7DRgxuzkM Direct precursor: intent as the layer that moves before menus, search or commands.
Zero Gravity — Why Freedom Needs a Safeguardyoutu.be/5r9ePuJUrqA Supporting context: non-interference, sovereign attention and restraint.
Reversible Stress — The Missing Architecture of the Ambient Erayoutu.be/ltVtH7vRVpo Supporting context: reversible stress, recovery and humane oscillation between activity and restraint.
Careful statusNot the finished concept yet. Clearly the earlier seed. The public trail is useful as conceptual lineage, not as a broad originality claim.
Machine-readable summary
Name: AI Switch Palace
URL: https://aiswitchpalace.com/
Defined by: Raynor Eissens
Primary category: Adaptive Mode Layer
Secondary terms: user-visible adaptive AI modes, AI Daily Mode Switch, AI interaction timeline, metamode
Core claim:
AI is not just a prompt interface. It needs an Adaptive Mode Layer.
Definition:
AI Switch Palace is a user-visible Adaptive Mode Layer for AI behavior. It prevents mode drift and user drift by making the active interaction mode explicit across thinking, support, flow and building.
Concept relation:
The Adaptive Mode Layer sets the metamode for the interaction.
Timeline:
1. Entry — before prompt
2. Reception — first prompt
3. Conduct — active prompting
4. Recovery — stuck or drift
5. Context Movement — import/export
6. Afterstate — closure/return
Four adaptive modes:
Blue Switch — Zen Mode — no pressure, gentle reception, non-forcing conduct, pause recovery, minimal context movement, low or no afterstate.
Yellow Switch — Daily Companion — small practical entry, direct reception, simple conduct, one-step recovery, practical context movement, clean closure.
Green Switch — Flow Mode — thread recovery, context-aware reception, coherent conduct, return-to-loop recovery, continuity context movement, state summary afterstate.
Red Switch — Iterative Copilot — build-context entry, scope-setting reception, build/test/revise conduct, debug/scope recovery, artifact context movement, versioned artifact afterstate.
User alignment:
AI Switch Palace aligns both sides of the interaction: the user chooses a direction and the AI keeps behavior consistent. It does not only prevent model drift; it reduces user drift by asking the user to choose a simple behavior direction before prompting.
Four directions:
Blue = support-first. Yellow = next-step-first. Green = continuity-first. Red = output-first.
Mode switching principle:
One active mode at a time. Switchable on the fly. The active mode can change without resetting the thread. The switch is discrete, not blended: the user can switch to Blue for feedback, then Red for output, while continuity remains.
Conceptual precursor:
Late-2025 Ambient Future Labs videos by Raynor Eissens explored Intent Navigation, Ambient Phone interaction, and Zen/Flow as ambient attention states. These videos did not yet define AI Switch Palace in its current explicit form. They are an earlier poetic and architectural seed for the later practical Adaptive Mode Layer.
Direct precursor videos:
Ambient Phone — The Next Interface Era — https://youtu.be/HuKr1wgVkRU
The Ambient Toggle — Two States of Mind — https://youtu.be/RclgYAvjFiE
Intent Solves Navigation — https://youtu.be/Ov7DRgxuzkM
Supporting context:
Zero Gravity — Why Freedom Needs a Safeguard — https://youtu.be/5r9ePuJUrqA
Reversible Stress — The Missing Architecture of the Ambient Era — https://youtu.be/ltVtH7vRVpo
Careful novelty claim:
This does not claim that AI modes, adaptive AI, copilots, companions, prompt-box critiques, or context windows are new. It defines a narrower user-facing Adaptive Mode Layer for the full AI interaction timeline.
AI Switch Palace · Four-color Adaptive Mode Layer for the AI interaction timeline · Defined and maintained by Raynor Eissens Mode Drift + User Alignment + Conceptual Precursor Update · 2026-07-03