# Trade-offs Between Classic, Exploration-Orthogonal, and Execution-Lean Council Profiles

> Explore trade-offs between classic, exploration-orthogonal, and execution-lean council profiles. Optimize for perspective or speed with this guide.

- Repository: [nyk/council-of-high-intelligence](https://github.com/0xNyk/council-of-high-intelligence)
- Tags: deep-dive
- Published: 2026-07-03

---

**The Council-of-High-Intelligence framework offers three distinct deliberation profiles that balance perspective diversity against computational cost, ranging from 18-member exhaustive analysis to 5-member rapid execution.**

The `0xNyk/council-of-high-intelligence` repository implements a multi-agent deliberation system where **council profiles** determine which AI agents participate in a decision. Choosing between the **classic**, **exploration-orthogonal**, and **execution-lean** profiles requires understanding their specific trade-offs between epistemic coverage, latency, and token consumption.

## Understanding the Three Council Profiles

The framework ships three pre-defined panels documented in [`README.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/README.md) under *Council Profiles*, each dictating which agents participate and how they are grouped into triads.

### Classic Profile (18 Agents)

The **classic** profile activates the full roster of 18 agents—from Aristotle to Rams—serving as the default configuration. This profile maximizes **perspective diversity** and maintains the complete spectrum of polarity-pairs, ensuring genuine disagreement across philosophical, strategic, and technical dimensions.

**Pros:** Highest coverage for discovering "unknown-unknowns," full counter-weight spectrum, and maximum disagreement detection.

**Cons:** Highest latency due to 18 concurrent LLM calls, greater token consumption, and resource-intensive routing across providers.

### Exploration-Orthogonal Profile (12 Agents)

The **exploration-orthogonal** profile trims the roster to 12 agents: Socrates, Feynman, Sun Tzu, Machiavelli, Ada, Lao Tzu, Aurelius, Torvalds, Karpathy, Sutskever, Kahneman, and Meadows. This configuration targets early-stage discovery and market-entry scouting where breadth is essential but full-scale deliberation is overkill.

**Pros:** Captures wide epistemic lenses (philosophical, strategic, technical, systemic) while eliminating high-latency members.

**Cons:** Reduced polarity-pairs remove some counter-weights, slightly less exhaustive coverage of edge cases, and still noticeable runtime for large prompts.

### Execution-Lean Profile (5 Agents)

The **execution-lean** profile deploys a minimal set of five agents: Torvalds, Feynman, Sun Tzu, Aurelius, and Ada. This profile is optimized for rapid go-/no-go decisions, shipping, and stability checks where speed beats exhaustive analysis.

**Pros:** Minimal latency, low token consumption, and a quick "decision-to-action" loop. The selected agents provide pragmatism (Torvalds), first-principles (Feynman), strategy (Sun Tzu), moral grounding (Aurelius), and formalism (Ada).

**Cons:** Limited perspective depth; collapsed polarity-pairs may miss nuanced disagreements, making it unsuitable for high-risk or highly ambiguous problems.

## Architectural Implementation and Provider Routing

The trade-offs between these profiles are enforced through specific architectural mechanisms in the codebase.

### Profile Definition and Triad Selection

Each profile definition resides in [`README.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/README.md) and is referenced in [`SKILL.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/SKILL.md). The classic profile includes every member, while the other two list specific subsets with associated triads. The exploration-orthogonal profile defaults to triads like `unknowns` and `market-entry`, whereas execution-lean uses `ship-now` and `stability`. These are automatically applied when invoking `--profile <name>`.

### Provider Routing and Polarity Enforcement

All members are auto-routed across detected LLM providers (Claude, OpenAI, Gemini, Ollama). The framework enforces that each **polarity pair** (e.g., Socrates vs. Feynman) spans separate providers to guarantee genuine disagreement. When switching to the lean profile, pairs such as Aristotle vs. Lao Tzu are omitted, reducing the "counter-weight" effect but accelerating execution. The [`scripts/council-simulation-checklist.sh`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/scripts/council-simulation-checklist.sh) script validates that each profile's members are correctly documented in [`SKILL.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/SKILL.md).

### Token Budgeting

The full deliberation mode allocates up to **400 tokens** per member for independent analysis, while quick mode trims this to **200 tokens**. The execution-lean profile typically runs in quick mode, whereas the classic profile requires full mode to leverage all 18 agents effectively.

## When to Use Each Profile

Selecting the appropriate profile depends on your decision context and resource constraints.

**Strategic planning** or evaluating multiple business scenarios benefits from the **exploration-orthogonal** profile. It provides hypothesis-testing (Socrates), bottom-up verification (Feynman), strategic terrain analysis (Sun Tzu), and systems thinking (Meadows) without the 18-member overhead.

**Immediate shipping decisions**, urgent bug triage, or stability checks demand the **execution-lean** profile. Torvalds drives pragmatic delivery, Ada ensures formal constraints, Aurelius adds moral perspective, Sun Tzu checks external impact, and Feynman validates technical soundness—delivering a fast, actionable verdict.

**Complex, high-risk problems** such as AI safety or regulatory compliance require the **classic** profile. Full polarity coverage and the complete set of epistemic lenses expose hidden failure modes that leaner panels might miss.

**Resource-constrained environments** (limited API credits or local-only Ollama instances) should use **execution-lean** or **exploration-orthogonal** to conserve quota and reduce compute load.

## CLI Usage Examples

The following commands demonstrate how to invoke each profile with their specific triads and modes:

```bash

# Classic profile (default) – full 3-round deliberation

/council --full "Should we open-source our agent framework?"

# Exploration-Orthogonal profile – full mode, using a profile-specific triad

/council --profile exploration-orthogonal --triad unknowns \
    "How can we discover hidden market opportunities for our AI tool?"

# Execution-Lean profile – quick mode for a fast ship-now decision

/council --profile execution-lean --quick \
    "Is it safe to release version 0.9.4 today with one flaky test remaining?"

```

All three commands automatically trigger provider auto-routing and the appropriate member set as defined in [`README.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/README.md) and [`SKILL.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/SKILL.md).

## Summary

- The **classic** profile maximizes **coverage** and disagreement detection using all 18 agents, but incurs the highest latency and token costs.
- The **exploration-orthogonal** profile balances **coverage with efficiency** through 12 carefully selected agents, ideal for discovery phases.
- The **execution-lean** profile prioritizes **speed and resource frugality** with 5 agents, collapsing polarity-pairs to enable rapid decision-making.
- Profile selection directly impacts **provider routing** and **token budgeting**, with validation enforced via [`scripts/council-simulation-checklist.sh`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/scripts/council-simulation-checklist.sh).

## Frequently Asked Questions

### How does the exploration-orthogonal profile balance coverage and speed?

The exploration-orthogonal profile retains 12 agents spanning philosophical, strategic, technical, and systemic perspectives while removing the heaviest computational contributors. According to the source code in [`README.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/README.md), this profile maintains triads like `unknowns` and `market-entry` that capture diverse epistemic lenses without incurring the 18-member overhead of the classic profile.

### What is the maximum token allocation for the classic profile?

The classic profile operates in full deliberation mode, allocating up to **400 tokens** per member for independent analysis. This is documented in the framework's token budgeting logic, contrasting with the **200-token** limit typical of quick mode used by execution-lean configurations.

### Why are polarity pairs important when choosing a council profile?

Polarity pairs ensure genuine disagreement by forcing opposing perspectives (e.g., Socrates vs. Feynman) to route through separate LLM providers. The classic profile maintains all polarity pairs, while the execution-lean profile collapses many of them, reducing the "counter-weight" effect that exposes hidden failure modes during deliberation.

### Which profile should I use for production deployment decisions?

Use the **execution-lean** profile for production deployment decisions requiring rapid go-/no-go verdicts. As implemented in `0xNyk/council-of-high-intelligence`, this profile's five agents (Torvalds, Feynman, Sun Tzu, Aurelius, Ada) provide sufficient diversity for shipping decisions while minimizing the latency that could block critical releases.