# Introducing Prime Agent: A self-improving RLM harness for coding and long-running …
Introduce product mechanisms through code and states, then conclude with task results and summary figures.
```motion-demo
{
"id": "sequence",
"title": "Overall choreography",
"src": "demos/sequence.html",
"duration": 44.44,
"status": "authored",
"description": "Motion and composition example",
"initialTime": 16,
"params": [
{
"key": "title",
"label": "Sample main copy",
"type": "text",
"value": "PRIME-AGENT",
"maxLength": 60
},
{
"key": "accent",
"label": "Accent color",
"type": "color",
"value": "#438bc3"
}
],
"presets": [
{
"label": "Default content",
"values": {
"title": "PRIME-AGENT",
"accent": "#438bc3"
}
},
{
"label": "OpenDesign adaptation",
"values": {
"title": "OpenDesign",
"accent": "#7050da"
}
}
]
}
```
## Keep granular backgrounds behind a readable terminal
Use dark particle terrain for depth and an opaque near-black terminal. Set headings, code, and logs in monospace, limiting blue-orange syntax colors to code and keeping backgrounds dimmer than text.
## Explain one relationship per mechanism page
Progress through tool calls, memory improvement, and agent communication. Match upper code with lower status structure, retaining subtask names across positions.
```motion-demo
{
"id": "clip-2",
"title": "Explain one relationship per mechanism page",
"src": "demos/clip-2.html",
"duration": 10,
"status": "authored",
"description": "10–20 seconds · Motion example",
"initialTime": 0.6
}
```
## Expand from one task to overall results
Show grid-task changes before gathering multiple results into a mosaic and enlarging the score, then move to summary curves. Hold numbers, task counts, and axis labels clearly.
| Time | Section |
|---|---|
| 0.00–1.80s | Black frame |
| 1.80–5.00s | Product title |
| 5.00–10.00s | Terminal mark |
| 10.00–16.00s | Parallel tool calls |
| 16.00–20.00s | Continuous improvement |
| 20.00–23.00s | Agent communication |
| 23.00–25.00s | Solving request |
| 25.00–28.00s | Grid task |
| 28.00–35.00s | Result mosaic and score |
| 35.00–38.00s | Curve comparison |
| 38.00–43.00s | Installation command |
| 43.00–44.44s | Brand |
## Replace code and results together with tasks
Adapt to coding, reasoning, or automation. Replace requests, code, task visuals, and scores together to explain one capability; give installation commands their own screen before the ending.
## Color and typography
Monospaced uppercase headings, commands, and code, with enlarged result figures.
#080807 · #eee9df · #438bc3 · #b09d7d · #d98442
## Elements and production rules
Granular terrain / terminal panel / subtask state / grid game / result score / curve chart
- Introduce product mechanisms through code and states, then conclude with task results and summary figures.
- Use dark particle terrain for depth and an opaque near-black terminal. Set headings, code, and logs in monospace, limiting blue-orange syntax colors to code and keeping backgrounds dimmer than text.
- Progress through tool calls, memory improvement, and agent communication. Match upper code with lower status structure, retaining subtask names across positions.
## Content replacement
- Adapt to coding, reasoning, or automation. Replace requests, code, task visuals, and scores together to explain one capability; give installation commands their own screen before the ending.
## Production parameters
[Visuals and objects](style.json) · [Section timing](timeline.json) · [Position data](tracks.json) · [File manifest](manifest.json)
This specification is based on the reference video. The HTML provides editable choreography examples; each file identifies the source, precision and valid interval of its data.
```motion-board
{
"version": 1,
"brand": "Prime Intellect",
"id": "V203",
"palette": [
"#080807",
"#eee9df",
"#438bc3",
"#b09d7d",
"#d98442"
],
"typography": "Uppercase monospace headings, commands and code, with prominently enlarged numeric results",
"sample": "PRIME-AGENT",
"objects": [
"Granular terrain",
"Terminal panel",
"Subtask state",
"Grid game",
"Result score",
"Curve chart"
],
"frames": [
{
"path": "assets/reference-1.jpg",
"timeSeconds": 0.0,
"frameId": "K0001",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-2.jpg",
"timeSeconds": 5.0,
"frameId": "K0010",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-3.jpg",
"timeSeconds": 16.0,
"frameId": "K0026",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-4.jpg",
"timeSeconds": 25.0,
"frameId": "K0040",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-5.jpg",
"timeSeconds": 35.0,
"frameId": "K0070",
"role": "reference-evidence",
"status": "measured"
},
{
"path": "assets/reference-6.jpg",
"timeSeconds": 43.0,
"frameId": "K0082",
"role": "reference-evidence",
"status": "measured"
}
],
"stages": [
{
"start": 0,
"label": "Black frame",
"end": 1.8,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 1.8,
"label": "Product title",
"end": 5,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 5,
"label": "Terminal mark",
"end": 10,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 10,
"label": "Parallel tool calls",
"end": 16,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 16,
"label": "Continuous improvement",
"end": 20,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 20,
"label": "Agent communication",
"end": 23,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 23,
"label": "Solving request",
"end": 25,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 25,
"label": "Grid task",
"end": 28,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 28,
"label": "Result mosaic and score",
"end": 35,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 35,
"label": "Curve comparison",
"end": 38,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 38,
"label": "Installation command",
"end": 43,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
},
{
"start": 43,
"label": "Brand",
"end": 44.44,
"status": "inferred",
"basis": "Representative-frame review; approximate boundaries"
}
],
"duration": 44.44,
"fontFamily": "ui-monospace, monospace"
}
```