Skip to content

TESSERACT.md

Latest commit

 

History

History
169 lines (122 loc) · 6.95 KB

TESSERACT.md

File metadata and controls

169 lines (122 loc) · 6.95 KB
Raw

TESSERACT Configuration

504 possible configurations for SPAR debates. Four axes, each independently selectable.

Version: 1.0 (Protocol v8.0)

Overview

TESSERACT is the configuration system for SPAR debates. By combining four independent axes (Depth, Pattern, Style, Horizon), practitioners can tune the debate format to match the decision context. The total configuration space is 3 × 6 × 7 × 4 = 504 unique combinations.

Most users never touch TESSERACT directly. The 5 presets cover 90% of use cases.

The Four Axes

1. Depth - How many voices?

Controls the number and diversity of personas in the arena.

Setting Agents Description
clash (default, minimum) 4 + PROBE NEWS compass (North, East, South, West) + framework critic.
rumble 8+ Full PERSONALITY roster + PROBE. Strategic pivots.
domain 4+N Clash + domain/MBS/governance experts.

2. Pattern - How do they interact?

Determines the conversation structure between personas.

Pattern Description Cultural Alias
sequential Each speaks in turn, building on previous Circle, Ubuntu
freeflow Unstructured, interrupt-driven -
binary Two sides, structured clash Oxford
dialectic (default) Thesis-antithesis-synthesis cycle Lincoln
inquisition One defends, others attack Scholastic
emergence Consensus-seeking through iterative refinement Nemawashi

3. Style - What's the tone?

Sets the argumentative posture of the debate.

Style Description
balanced (default) Even-handed exploration
adversarial Attack the position. Find weaknesses.
steelman Each persona must strengthen the opposing argument before attacking
consensus Build agreement progressively
premortem Assume the decision failed. Work backwards to find why.
escalation Start mild, increase pressure each round
inversion Argue the opposite of what you believe

4. Horizon - What timeframe?

Anchors the debate to a specific temporal lens.

Horizon Timeframe Question
now Days to weeks "What do we do Monday?"
cycle (default) Quarter to year "What's the right call this cycle?"
legacy Multi-year "How does this look in 5 years?"
shifted Radically different future "If everything changed, would this still hold?"

5. Width - How deep per response? (token-based SPARs only)

Controls the token budget per persona per round. Applies only to LLM-executed SPARs, not human facilitation.

Width Tokens/response Use Case
quick ~200-400 Fast signal, rapid iteration. Headlines only.
standard (default) ~600-1000 Balanced depth. Evidence + reasoning.
ultra ~1500-2500 Full analysis. Citations, worked examples, deep chains.

Note: Width does not multiply the 504 base configurations. It is a runtime parameter, not a design axis. The 504 count (3 × 6 × 7 × 4) covers the four design axes: Depth × Pattern × Style × Horizon.

Presets

Presets are curated TESSERACT configurations for common scenarios.

Preset Depth Pattern Style Horizon Width Use Case
(default) clash dialectic balanced cycle standard General decision-making
⚡ quick clash binary balanced now quick Fast yes/no decisions
🔮 deep rumble dialectic steelman legacy ultra Strategic pivots, M&A
🚨 crisis clash freeflow escalation now standard Surface hidden issues fast
🤝 align clash sequential consensus cycle standard Build team agreement
🎯 challenge clash binary adversarial now standard Stress-test a specific idea

Agent Roster

Clash (4 + Centre)

Agent Direction Element Core Question
💨 Visionary North Air "Where are we going?"
🔥 Challenger East Fire "What's emerging?"
🌍 Pragmatist South Earth "What's grounded?"
💧 Sage West Water "What has history taught?"
👁️ Observer Centre-OUTER Soul "Why will humans do or not do this?"
🔍 PROBE Centre-INNER Soul "Is the way we're reasoning flawed?"

Centre is always-on. OUTER (Observer, Human 5 Whats) looks at human behavior. INNER (PROBE, 5-point critique) looks at the methodology. Both run in every debate. See CENTRE.md.

Rumble adds (8+)

The PERSONALITY framework provides 11 additional archetypes:

Letter Archetype Core Question
P Protector "What could go wrong?"
E Explorer "What's possible?"
R Realist "What's feasible?"
S Strategist "What are the ripples?"
O Observer "What's the pattern?"
N Navigator "Where should we head?"
A Advocate "Who's not being heard?"
L Literary "What does wisdom say?"
I Integrator "How do we synthesize this?"
T Timekeeper "What does the timeline tell us?"
Y Yielder "What should we let go?"

Combining Axes

You can mix any combination:

# Steelman debate with 8 agents, looking at multi-year impact
sparkit --rumble --style=steelman --horizon=legacy "Should we open-source our core engine?"

# Quick adversarial check on a tactical decision
sparkit --style=adversarial --horizon=now "Should we ship this feature Friday?"

# Consensus-building with the full team
sparkit --pattern=sequential --style=consensus "What should our Q2 priorities be?"

Choosing Your Configuration

                 ┌─────────────────────────────────────────┐
                 │ What kind of decision is this?          │
                 └─────────────┬───────────────────────────┘
                               │
              ┌────────────────┼────────────────┐
              │                │                │
         Quick call?      Strategic?       Team alignment?
              │                │                │
         ⚡ --quick       🔮 --deep         🤝 --align
              │                │                │
        Need to stress    Crisis mode?    Adversarial test?
        test further?         │                │
              │           🚨 --crisis     🎯 --challenge
         Custom TESSERACT

See also: PROTOCOL.md | DEPTH_MODES.md | PERSONA_LIBRARY.md