v1.0.0
NationalSecurityAgency/ghidrav1.0.0Mar 30, 2026by 0xNyk
Upgrading?
AI Summary
The first stable release of Council of High Intelligence, a deliberation tool in which 18 historical-thinker personas analyze hard problems. This version adds multi-provider auto-routing so council members are distributed across Anthropic Claude, OpenAI Codex, Google Gemini, and local Ollama models for genuine model diversity with zero configuration. It also introduces a problem restate gate, reorders verdict templates to surface actionable output first, and streamlines the enforcement protocol from nine steps to seven.
Key Highlights
- Multi-provider auto-routing with zero config: auto-detects installed LLM providers and spreads council members across them, separating polarity pairs and respecting per-member affinity hints
- New problem restate gate forces each member to restate the problem and offer an alternative framing before analysis, catching wrong-question failures early
- Verdict templates reordered to lead with Unresolved Questions and Recommended Next Steps instead of burying them under epistemic scorecards
- Protocol streamlined from 9 enforcement steps to 7 via a consolidated post-round scan, plus outcome tracking added to all verdict templates
- Ships with 18 defined council members across domains (Aristotle, Feynman, Kahneman, Taleb, Meadows, etc.) and 3 deliberation modes: Full, Quick, and Duo
Breaking Changes
- Verdict output format changed: Unresolved Questions and Recommended Next Steps now appear first, ahead of epistemic scorecards, which may break downstream parsing or reader expectations
- Member metadata consolidated into agent YAML frontmatter as the single source of truth, replacing prior separate metadata definitions
- Enforcement protocol restructured from 9 steps to 7, so any custom prompts or tooling referencing the old step sequence will need updating
New Features
- Multi-provider auto-routing across Anthropic Claude, OpenAI Codex, Google Gemini, and Ollama local models
- Auto-detection script that probes for available providers and emits structured JSON
- Routing algorithm that separates polarity pairs across providers, spreads members evenly, and honors per-member affinity hints
- Dispatch logic for 'codex exec', 'gemini -p', and 'ollama run' with automatic fallback to Claude on failure
- New CLI flags: --no-auto-route and --dry-route
- Problem Restate Gate requiring problem restatement and alternative framing before analysis
- Verdict template reorder prioritizing Unresolved Questions and Recommended Next Steps
- Outcome tracking via a Follow-Up section in all verdict templates
- Consolidated member metadata in agent YAML frontmatter
- Three deliberation modes: Full (3-round with cross-examination and synthesis), Quick (2-round), and Duo (2-member polarity-pair dialectic)
- 18-member council roster with per-member domain assignments and default model mappings
Full Release Notes
# Council of High Intelligence v1.0.0 The first stable release. 18 historical thinkers deliberate on your hardest problems — now across multiple AI providers. ## What's New ### Multi-Provider Auto-Routing The council automatically detects installed LLM providers (Anthropic Claude, OpenAI via Codex, Google Gemini, Ollama local models) and distributes members across them for genuine model diversity. Zero config required. - **Auto-detection script** probes for available providers and outputs structured JSON - **Routing algorithm** separates polarity pairs across providers, spreads evenly, respects per-member affinity hints - **Dispatch logic** for `codex exec`, `gemini -p`, `ollama run` with automatic fallback to Claude on failure - New flags: `--no-auto-route`, `--dry-route` ### Problem Restate Gate Each council member must restate the problem + provide an alternative framing before analysis begins. Catches wrong-question failures before burning rounds on them. ### Verdict Template Reorder Verdicts now lead with **Unresolved Questions** and **Recommended Next Steps** — what the user needs to act on — instead of burying them below epistemic scorecards. ### Streamlined Protocol - Merged 3 enforcement steps into 1 post-round scan (9 steps → 7) - Added outcome tracking (Follow-Up section in all verdict templates) - Consolidated member metadata into agent YAML frontmatter (one source of truth per member) ## The 18 Council Members | Member | Domain | Default Model | |--------|--------|---------------| | Aristotle | Categorization & structure | opus | | Socrates | Assumption destruction | opus | | Sun Tzu | Adversarial strategy | sonnet | | Ada Lovelace | Formal systems & abstraction | sonnet | | Marcus Aurelius | Resilience & moral clarity | opus | | Machiavelli | Power dynamics & realpolitik | sonnet | | Lao Tzu | Non-action & emergence | opus | | Feynman | First-principles debugging | sonnet | | Linus Torvalds | Pragmatic engineering | sonnet | | Miyamoto Musashi | Strategic timing | sonnet | | Alan Watts | Perspective & reframing | opus | | Andrej Karpathy | Neural network intuition | sonnet | | Ilya Sutskever | Scaling frontier & AI safety | opus | | Daniel Kahneman | Cognitive bias & decision science | opus | | Donella Meadows | Systems thinking & feedback loops | sonnet | | Charlie Munger | Multi-model reasoning & economics | sonnet | | Nassim Taleb | Antifragility & tail risk | opus | | Dieter Rams | User-centered design | sonnet | ## 3 Deliberation Modes - **Full** — 3-round deliberation with cross-examination, enforcement, and synthesis - **Quick** — 2-round rapid analysis for simpler questions - **Duo** — 2-member dialectic using polarity pairs ## Installation ```bash git clone https://github.com/0xNyk/council-of-high-intelligence.git cd council-of-high-intelligence ./install.sh ``` Then restart Claude Code and use `/council` to convene.
Related Repositories
krahets/hello-algo★ 130,899
《Hello 算法》：动画图解、一键运行的数据结构与算法教程。支持简中、繁中、English、日本語，提供 Python, Java, C++, C, C#, JS, Go, Swift, Rust, Ruby, Kotlin, TS, Dart 等代码实现
iluwatar/java-design-patterns★ 94,813
Design patterns implemented in Java
OtterMind/Chat2DB★ 28,329
Chat2DB is a free, cross-platform, local-first database client and SQL workspace for developers, DBAs, analysts, and data teams. Connect to 40+ databases, manage data, edit and run SQL, and use your own AI model to generate, explain, and optimize queries. Available on desktop, web, Docker, and CLI, with MCP support.