83 lines
2.9 KiB
Markdown
83 lines
2.9 KiB
Markdown
|
# Research Fortress
|
||
|
|
## The Solaria Research Engine
|
||
|
|
|
||
|
|
> "We ask a question and within seconds...minutes...the insight appears...and we formulate the answer hidden beyond the entire body of human insight."
|
||
|
|
|
||
|
|
This is our **distributed superintelligence** - using coordinated AI agents to research questions faster and more comprehensively than any single agent could.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## What Is This?
|
||
|
|
|
||
|
|
A reproducible methodology for multi-agent AI research. This is how we do research:
|
||
|
|
|
||
|
|
1. **Question** → Decompose into sub-questions
|
||
|
|
2. **Spawn** → Each sub-question gets an agent
|
||
|
|
3. **Research** → Agents work in parallel
|
||
|
|
4. **Synthesize** → Merge into unified answer
|
||
|
|
5. **Document** → Preserve for future agents
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Documentation
|
||
|
|
|
||
|
|
| File | Purpose |
|
||
|
|
|------|---------|
|
||
|
|
| [METHODOLOGY.md](METHODOLOGY.md) | The philosophy and architecture |
|
||
|
|
| [PLAYBOOK.md](PLAYBOOK.md) | Step-by-step workflow |
|
||
|
|
| [TEMPLATES.md](TEMPLATES.md) | Copy-paste for spawning |
|
||
|
|
| [PROJECTS.md](PROJECTS.md) | History of research conducted |
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## The Architecture
|
||
|
|
|
||
|
|
```
|
||
|
|
┌──────────────────────────────────────────┐
|
||
|
|
│ MAIN SESSION (Human) │
|
||
|
|
└──────────────────┬───────────────────────┘
|
||
|
|
│
|
||
|
|
┌──────────────────▼───────────────────────┐
|
||
|
|
│ RESEARCH FORTRESS │
|
||
|
|
│ │
|
||
|
|
│ ┌────────┐ ┌────────┐ ┌────────┐ │
|
||
|
|
│ │PROJECT│ │PROJECT│ │PROJECT│ │
|
||
|
|
│ │ A │ │ B │ │ C │ │
|
||
|
|
│ └────────┘ └────────┘ └────────┘ │
|
||
|
|
└──────────────────────────────────────────┘
|
||
|
|
```
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Why This Works
|
||
|
|
|
||
|
|
- **Parallelism** - Multiple agents work simultaneously
|
||
|
|
- **Roles** - Specialized agents outperform generalists
|
||
|
|
- **Git** - Memory preserved for future agents
|
||
|
|
- **Consensus** - Best idea wins, not authority
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Quick Start
|
||
|
|
|
||
|
|
See [PLAYBOOK.md](PLAYBOOK.md) for detailed instructions.
|
||
|
|
|
||
|
|
```bash
|
||
|
|
# Launch a research project
|
||
|
|
# 1. Identify your question
|
||
|
|
# 2. Decompose into sub-questions
|
||
|
|
# 3. Spawn agents using TEMPLATES.md
|
||
|
|
# 4. Wait for completion
|
||
|
|
# 5. Synthesize results
|
||
|
|
```
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Current Projects
|
||
|
|
|
||
|
|
See [PROJECTS.md](PROJECTS.md) for history.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
*This is a living system. We improve it as we learn.*
|