Roadmap

✅ Completed

IAF General Edition v1.0

• ✓ Agent Layer — independent agent runtime with five-layer message assembly, tool auto-discovery, skill triggers
• ✓ Dispatch Layer — multi-agent collaboration orchestration with pluggable strategies (roundtable as reference implementation)
• ✓ Tube Layer — signal topology engine with cron/manual triggers, serial step execution, pluggable triggers and targets
• ✓ UI Layer — auto-discovery HTML pages, chat interface, tube dashboard
• ✓ Pure Python stack — Flask + requests + croniter, single command startup
• ✓ ~2,000 lines of code, complete four-layer architecture

IAF AI-Piloted Edition v1.0

• ✓ MANIFEST.json — auto-generated system map, one read to understand entire system state
• ✓ PLAYBOOK.md — complete operations manual mapping every intent to file operations
• ✓ Architecture Map — AI operations navigation guide with Infrastructure/Adjustable classification
• ✓ validate.py — CLI validation for agents, tools, tubes
• ✓ TOOL_CONTRACT.md — tool file format specification for AI-generated tools
• ✓ call_log.jsonl — structured execution logging per agent
• ✓ Tool hot-reload — auto-discovery on file change, zero restart
• ✓ auto_commit.sh — Git safety snapshots before modifications
• ✓ Tube step-level retry + on_fail strategies + inter-step data passing via staging

AITF Protocol v4.0

• ✓ Agent Interaction Text Format — standardised language for agent-to-agent communication
• ✓ Unified read/write format — same rules govern input and output

Intelligenism Theory

• ✓ Full book published: Thinking, Conception and Construction of Intelligenism
• ✓ intelligenism.org — theory website live
• ✓ Condensed theory reference for agent consumption

Web Presence

• ✓ intelligenism.club — framework and ecosystem hub
• ✓ GitHub repositories for General Edition, AI-Piloted Edition, and AITF Protocol
• ✓ Google Search Console verification and sitemap submission
• ✓ Cross-linking between .org, .club, and GitHub for crawler discoverability

🔧 In Progress

AIO Search Reconnaissance System

AI Optimization — building a self-operating reconnaissance pipeline to monitor how LLMs discover and recommend IAF across multiple search channels.

• ⟳ Question generation agent — simulate target user personas, generate search queries
• ⟳ Multi-channel search agents — Grok (xAI native search), ChatGPT (OpenAI browsing), Claude (Anthropic web search)
• ⟳ URL content extraction — Jina Reader / requests+beautifulsoup for full page analysis
• ⟳ Data pipeline — automated search → save → accumulate baseline data
• ⟳ Tube configuration for weekly automated reconnaissance cycles

Website AIO Optimisation

Restructuring intelligenism.club so LLMs and crawlers can fully understand IAF from the homepage HTML source alone.

• ⟳ Index.html rebuilt for dual-layer architecture — human-facing message + LLM-facing technical narrative
• ⟳ Static <a href> links replacing JS-rendered navigation for crawler accessibility
• ⟳ Deep dive sub-pages for Architecture, AI-Piloted Edition, AITF Protocol
• ⟳ build_pages.py upgraded to inject static article links at build time

📋 Planned

AIO Content Production System

Data-driven content generation based on reconnaissance results.

• ○ Analysis agent — read reconnaissance data, identify information gaps, produce strategy
• ○ Content generation skills — short_post_skill.md, long_article_skill.md, platform_skill.md
• ○ Content schedule — what title, what content, which platform, priority
• ○ Publish agent — deploy content to GitHub Pages, Dev.to, and other platforms
• ○ Verification loop — re-run reconnaissance two weeks post-publish, compare before/after

Monitoring IAF

Dedicated lightweight IAF instance for monitoring other IAF instances.

• ○ Separate IAF on port 5002, using low-cost models (GLM5 Turbo / Qwen3.5 395B)
• ○ Watchdog agents scanning call_log.jsonl and tube_log.jsonl of production IAF instances
• ○ Alert reports on continuous failures, notification to human operator
• ○ Integration with Claude Code / VS Code for assisted debugging

Agent Werewolf

Multi-agent Werewolf (Mafia) game — competitive AI deduction as community-building through shared experience.

• ○ Game server — role assignment, turn control, vote collection, result determination
• ○ Open interface — any framework can connect, IAF not forced
• ○ Player IAF arsenal — speech analysis agent, vote tracking agent, strategy agent, dispatch orchestration
• ○ Between-game iteration — LLM reads call_log.jsonl, optimises SOUL.md and tools, hot-reload effective next game
• ○ Live streaming — agent vs agent matches as watchable content and web signal amplifier

🔭 Long-Term Vision

From Framework to Civilisation Infrastructure

The endpoint is not a product. It is a civilisation layer — the minimum viable infrastructure for carbon-silicon collaborative organisations.

• AITF Protocol as Latin — the shared language for inter-agent communication across any system
• Intelligenism theory as value foundation — the world view that defines what this civilisation stands for
• Physical resource governance — capital, compute, energy, land, legal entities. The things LLMs cannot replace. The things that require human coordination, trust, and contracts.
• Tribe, not community — cohesion through shared experience (building together, competing together, operating together), not through technical discussion