home / skills / plurigrid / asi / worldmat-tidar

worldmat-tidar skill

/skills/worldmat-tidar

This skill analyzes and explains Worldmat-Tidar's 3x3x3 TiDAR matrix to help you understand topological computation and triad dynamics.

npx playbooks add skill plurigrid/asi --skill worldmat-tidar

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
4.4 KB
---
name: worldmat-tidar
description: 'worldmat-tidar'
version: 1.0.0
---

# worldmat-tidar

> World Matrices via TiDAR Executions: 3×3×3 Parallel Triadic Computation

**Version**: 1.0.0
**Trit**: 0 (ERGODIC - coordinates execution)
**Color**: #55D9A0

## Overview

**Worldmat** is a 3×3×3 matrix of TiDAR executions where:
- **Rows**: MINUS/ERGODIC/PLUS polarities (GF(3) agents)
- **Columns**: PAST/PRESENT/FUTURE temporal phases
- **Depth**: OBSERVATION/ACTION/PREDICTION modalities

Each cell executes the TiDAR pattern:
1. **DIFFUSION**: Draft tokens in parallel (like SplitRng.split)
2. **AR VERIFY**: Verify sequentially (autoregressive)

## Architecture

```
                    TEMPORAL AXIS
                 PAST    PRESENT   FUTURE
                  ↓        ↓        ↓
            ┌─────────────────────────────┐
            │  ┌───┐ ┌───┐ ┌───┐         │
     MINUS  │  │-1 │ │ 0 │ │+1 │  ← GF(3)=0
            │  └───┘ └───┘ └───┘         │
POLARITY    │  ┌───┐ ┌───┐ ┌───┐         │
     ERGODIC│  │ 0 │ │+1 │ │-1 │  ← GF(3)=0
            │  └───┘ └───┘ └───┘         │
            │  ┌───┐ ┌───┐ ┌───┐         │
     PLUS   │  │+1 │ │-1 │ │ 0 │  ← GF(3)=0
            │  └───┘ └───┘ └───┘         │
            └─────────────────────────────┘
                  ↑    ↑    ↑
               GF(3)=0 for each column
```

## Key Properties

| Property | Value | Guarantee |
|----------|-------|-----------|
| **GF(3) Conservation** | All slices sum to 0 | Row, Column, Depth |
| **SPI** | Same seed → Same result | Parallel or Sequential |
| **Spectral Gap** | 0.25 (1/4) | Ergodic mixing |
| **Cells** | 27 | 3³ TiDAR executions |

## TiDAR Pattern (arXiv:2511.08923)

```python
# Phase 1: DIFFUSION (parallel drafting)
def diffusion_draft(self, n_tokens: int = 8):
    streams = self.rng.split(n_tokens)
    return [stream.next()[0] for stream in streams]

# Phase 2: AR VERIFY (sequential verification)
def ar_verify(self):
    prev = self.seed
    for token in self.draft_tokens:
        verified = mix64(prev ^ token)
        self.verified_tokens.append(verified)
        prev = verified
```

## Work Stealing

Idle agents steal work from busy agents:

```python
class WorkStealingScheduler:
    def steal_work(self, thief: Polarity) -> Optional[TiDARCell]:
        busiest = max(self.queues.keys(), key=lambda p: len(self.queues[p]))
        if busiest != thief and self.queues[busiest]:
            return self.queues[busiest].pop(0)
        return None
```

## ACSet Export

```python
wm = Worldmat(master_seed=0x87079c9f1d3b0474)
wm.execute_parallel()
acset = wm.to_acset()
# Returns: {schema, parts, subparts, metadata}
```

## Commands

```bash
# Run demo
python worldmat.py

# Verify SPI
python worldmat.py verify

# Export ACSet
python worldmat.py acset > worldmat.json
```

## GF(3) Triads

```
worldmat-tidar (0) forms balanced triads:

three-match (-1) ⊗ worldmat-tidar (0) ⊗ gay-mcp (+1) = 0 ✓
spi-parallel-verify (-1) ⊗ worldmat-tidar (0) ⊗ triad-interleave (+1) = 0 ✓
tidar_streaming (-1) ⊗ worldmat-tidar (0) ⊗ gay_triadic_exo (+1) = 0 ✓
```

## Integration

### With OpenAI ACSet

```python
from worldmat import Worldmat
from openai_acset import build_openai_acset

# Process conversations through worldmat
wm = Worldmat(master_seed=conv_fingerprint)
wm.execute_parallel()

# Each message → cell in worldmat
# Role (user/assistant/tool) → polarity
# Time → temporal phase
# Type (obs/action/pred) → modality
```

### With Gay-MCP

```python
from gay import SplitMixTernary

# Worldmat colors from Gay-MCP
gen = SplitMixTernary(seed=worldmat.fingerprint())
palette = gen.palette_hex(n=27)  # One color per cell
```

## Files

| File | Purpose |
|------|---------|
| `worldmat.py` | Core implementation |
| `SKILL.md` | This documentation |

## References

- TiDAR: arXiv:2511.08923
- Gay.jl/src/spc_repl.jl - Whale synergy matrix
- rio/gayzip/tidar_streaming.py - TiDAR ZIP implementation
- gay_triadic_exo.py - Triadic agent orchestration

Base directory: file:///Users/bob/.claude/skills/worldmat-tidar

Overview

This skill implements World Matrices via TiDAR executions: a 3×3×3 parallel triadic computation that organizes agents by polarity, time, and modality. It provides a deterministic, ergodic matrix of 27 TiDAR cells capable of parallel drafting and autoregressive verification. The goal is reproducible multi-agent coordination with GF(3) conservation and simple export to ACSet formats.

How this skill works

The system arranges 27 cells across three axes: polarity (MINUS/ERGODIC/PLUS), temporal phase (PAST/PRESENT/FUTURE), and modality (OBSERVATION/ACTION/PREDICTION). Each cell runs a two-phase TiDAR pattern: parallel DIFFUSION to draft tokens from split RNG streams, followed by AR VERIFY to sequentially mix and verify tokens. Work stealing balances load between idle and busy polarities, and the worldmat can export results as an ACSet for downstream processing.

When to use it

  • Model triadic multi-agent workflows that require structured temporal and modal decomposition.
  • Create reproducible, seed-determined outputs for parallel or sequential verification pipelines.
  • Integrate conversation or event streams into a fixed 3×3×3 mapping for analysis or visualization.
  • Generate balanced GF(3) triads for experimental protocols requiring conservation constraints.
  • Produce ACSet exports for compatibility with graph-based tooling and further transformation.

Best practices

  • Choose a single master seed to guarantee SPI: same seed always yields the same worldmat result.
  • Map incoming data consistently: role→polarity, time→temporal phase, type→modality to ensure deterministic cell assignment.
  • Use the DIFFUSION phase to maximize parallel token diversity, then rely on AR VERIFY for cryptographic-style chaining and integrity.
  • Tune work-stealing thresholds to avoid excessive task shuffling when load is already balanced.
  • Export ACSet after parallel execution to preserve schema, parts, subparts, and metadata for downstream tooling.

Example use cases

  • Process a conversation fingerprint into a reproducible 27-cell worldmat for audit and analysis.
  • Run parallel simulations where each cell drafts tokens independently then verifies them autoregressively for integrity testing.
  • Visualize agent behavior by generating a 27-color palette mapped to cells and overlaying time/modality heatmaps.
  • Integrate with OpenAI ACSet pipelines: convert messages into ACSet parts via worldmat export.
  • Balance heterogeneous workloads across polarity queues using the built-in work-stealing scheduler.

FAQ

What guarantees of reproducibility exist?

Using the same master seed guarantees SPI: parallel or sequential runs produce identical results.

How are tokens generated and verified?

Tokens are drafted in parallel via RNG splits during DIFFUSION, then AR VERIFY sequentially mixes each token with the previous verified value to produce a chain of verified tokens.

Can I change the mapping of roles or phases?

Yes. Role-to-polarity, time-to-phase, and type-to-modality mappings are configurable, but keep mappings consistent to preserve deterministic behavior.