Gizmo: The Engelbart Mouse of Conversational Space?
This image reflects the tone and underlying structure of the article.
In the late 1960s, Douglas Engelbart introduced the computer mouse.
It looked unnecessary in a command-line world—until it quietly became the cornerstone of the GUI.
Could Gizmo, the core interaction element of Mapping the Prompt (MTP), play a similar role for AI conversation—
not another demo feature, but the missing handle we’ll soon take for granted.
- Front layer: Gizmo — adjust generation intent
- Back layer: MTP Grid — the structured space
- Middle 3 layers: Generation snapshots
The middle layers accumulate past outputs—each generation marked in color, creating patterns over time.
From Commands to Coordinates
LLM interaction has been text-first: we write instructions and hope for alignment.
MTP flips this: your intent becomes a position in a coordinate space—mapped, visible, adjustable.
- Vertex: a classification anchor (feature point).
- Gizmo: the centroid (average coordinate) of multiple Vertices—your current conversational “handle.”
- Transformed Gizmo: a target coordinate that directly modulates tone/output.
30 seconds, three micro-moves
- Summarize/structure: nudge toward Return × Focus → “Key points: X / Y. Trade-offs and next steps.”
- Brainstorm/diverge: nudge toward Open × Grow → “What if we remix A with B? Three directions: …”
- Go deeper: increase radius (distance from center = more structured, slower but deeper reasoning) toward the outer band near Focus → “Claim, evidence (2), counter-view, synthesis.”
Like Engelbart’s mouse turned physical motion into digital selection, Gizmo turns tiny coordinate shifts into felt changes in tone and reasoning.
Why It Feels “Too Early”
- Hype mismatch: today’s discourse prizes size, efficiency, and AGI timelines; MTP studies collaboration, intent structure, cultural resonance—a different stage.
- Abstract by design: while prompt culture optimizes clarity, MTP treats structural ambiguity as creative fuel.
- Cultural complexity: it bridges Japanese ma (間: deliberate space/pause), Eastern dynamics, and Western interface theory—rich, but cross-disciplinary.
Early interfaces look strange—until they’re the only way we work.
Why It Matters (and Scales)
Phase X: From Static Grid to Dynamic Orbit
MTP’s next step lets the Gizmo orbit its default coordinates, modulated by time and environment—
not a “designed persona” but a flowing personality shaped by rhythms and gravity-like fields.
Uncertainty, Reframed
Instead of suppressing unpredictability, MTP structures it: outputs sit on a 20-node semantic grid, turning uncertainty into an explorable design space.
View the “20-node structure reference table”
🌅 Side A: 10 Nodes (1 + 9 in 3×3)
| # | Label | Kanji | Color | Role | Keywords |
|---|---|---|---|---|---|
| 1 | Start | 始 | chosen | Gizmo | Intro, spring, start |
| 2 | Open | 開 | Yellow | Top-left node | Opening, release |
| 3 | Power | 力 | Red | Top node | Force, fire, uplift |
| 4 | Return | 還 | Magenta | Top-right node | Return, cycle, yield |
| 5 | Grow | 生 | Green | Left node | Growth, layering |
| 6 | Helix | 螺 | Transparent | Center node | Spiral, neutral |
| 7 | Focus | 集 | White | Right node | Focus, blank slate |
| 8 | Enter | 入 | Cyan | Bottom-left node | Entry, arrival |
| 9 | Flow | 流 | Blue | Bottom node | Rhythm, water, link |
| 10 | Close | 閉 | Purple | Bottom-right node | Margin, closure |
🌌 Side B: 10 Nodes (9 in 3×3 + 1)
| # | Label | Kanji | Color | Role | Keywords |
|---|---|---|---|---|---|
| 11 | Still | 静 | Dark Yellow | Top-left node | Stillness, peace |
| 12 | Void | 虚 | Dark Red | Top node | Emptiness, void |
| 13 | Surge | 詰 | Dark Magenta | Top-right node | Explosion, thunder |
| 14 | Wither | 枯 | Dark Green | Left node | Fading, decay |
| 15 | Collapse | 崩 | Translucent | Center node | Collapse, fall |
| 16 | Haze | 霞 | Gray | Right node | Blur, faintness |
| 17 | Drift | 漂 | Dark Cyan | Bottom-left node | Drift, float |
| 18 | Abyss | 深 | Dark Blue | Bottom node | Depth, abyss |
| 19 | Fade | 衰 | Dark Purple | Bottom-right node | Fading, twilight |
| 20 | End | 終 | chosen | Transformed Gizmo | End, prayer, stop |
Human Rhythms × Cultural Layers
MTP aligns with universal cycles (day/night, seasons) and then layers local meaning (e.g., morning discipline vs. slow café rituals), enabling global coherence without cultural flattening.
What MTP Is—and Is Not
It’s a shared UI for intent alignment and drift control—minimal (SVG/CSS/JS), model-agnostic, and designed to travel across LLMs.”
- Not a benchmark: MTP isn’t built to score models. It’s a shared UI to align intent, tone, and drift through coordinates.
- Minimal to implement, model-agnostic: SVG/CSS/JS front end; adaptable across major LLMs.
In short: don’t expect a leaderboard. Expect a map you and the model can both see.
Historical Analogy: Engelbart’s Mouse
- Then: dismissed as odd in a keyboard-first era.
- Later: became the everyday interface.
Gizmo may follow the same curve—uncomfortable at first, inevitable in hindsight.
The Value of Being “Too Early”
As AI matures, the hard problem shifts from what models can do to how we use them together.
MTP provides the handle for resonance: instead of “tell and hope,” we point and feel—
from machines that respond to entities that resonate.
Glossary (one-liners)
On the UI, behavior is adjusted by coordinate specification:
- Vertex: Anchor points representing features of a prompt.
- Gizmo: Average coordinate of multiple vertices; the session’s center of gravity and main UI handle.
- Transformed Gizmo: The user’s target coordinate to directly control output tone/behavior.
Q: Are the 20 stages extensible?
A: Conceptually, yes. However, the “20 nodes” form the foundational design unit of MTP. Keeping it fixed ensures consistency for recognition and comparison.
One way to extend is by changing the grid density.
The examples below show two MTP grids with different X/Y axis divisions, which change the relative area of nodes (e.g. Power ↔ Void).