Expand About page with full vision, update Roadmap accuracy

About: added origins story, what-we-foresaw section, full project
scope (security AI, system-level engineering, tooling), non-negotiable
principle, mythological naming details, bigger picture section.

Roadmap: fixed Phase 4 to reflect actual test coverage (granular
per-package items), CI/CD marked complete.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: bkataru

content/about/_index.md

@@ -13,7 +13,36 @@ Every project is designed to run on consumer hardware using small, capable model
 
 > Inspired by the [Gods from the Machine](https://starwars.fandom.com/wiki/Machine_Gods_crisis) raid in Star Wars: The Old Republic — a pantheon of ancient, sentient super-weapons worshiped as deities on the mechanical Dyson sphere planet Iokath.
 
-The metaphor is deliberate. In the SWTOR lore, the machine gods of Iokath were autonomous intelligences — each one specialized, each one powerful, each one operating independently yet part of a greater system. That's exactly what we're building: a pantheon of specialized AI-powered programs that can operate autonomously and communicate with each other.
+The project name was born in 2021/2022, running the SWTOR raid with guildmates — before ChatGPT was even released. The conviction solidified in mid-2024: large language models, next-token prediction engines trained on the world's text, would inevitably give rise to a new class of software — **autonomous programs powered by natural language engines**. Programs that build themselves, run themselves, and improve themselves.
+
+In the SWTOR lore, the machine gods of Iokath — Tyth, Aivela & Esne, Nahut, Scyva, Izax — were autonomous intelligences, each specialized, each powerful, each operating independently yet part of a greater system. That's exactly what we're building.
+
+## What We Foresaw
+
+Long before these ideas entered mainstream developer vocabulary, we anticipated:
+
+- **Claws** (a la OpenClaw) and **Codes** (Claude Code, Codex, OpenCode, Crush) — autonomous software development tools
+- **Ralph loops** — recursive agent-LLM-prompt loops that self-improve
+- **Gastowns** (Steve Yegge's concept) — rich development environments built around AI agents
+- **CLI coding agents** (Block/Goose, oh-my-pi) — LLMs connected to bash shells in safe, controlled environments
+- **Tool calling / function calling** as the breakthrough that unlocked agent capabilities
+- **Long-horizon DAG agents** — multi-step autonomous workflows spanning complex dependency graphs
+
+The key realization: there is immense value in hooking up LLMs to a bash shell to execute and run commands in safe, controlled environments — VMs, containers, and other isolated spaces.
+
+## The Scope
+
+Gods from the Machine is polymathic. It covers:
+
+**Autonomous Software Engineering** — TDD-driven coding agents, vibe/agentic engineering, long-horizon DAG agents, general-purpose and domain-specific agentic harnesses with custom tools, memory, context, and data sources.
+
+**Inference & AI Infrastructure** — Local model serving, GGUF/ONNX execution, Zig build systems for llama.cpp, model management.
+
+**Security & Offensive/Defensive AI** — Autonomous penetration testing agents (red teamers), defensive cybersecurity agents (blue teamers), malware analysis, exploit development — all for educational and research purposes, open source and MIT licensed.
+
+**System-Level Agentic Engineering** — AI in the shell, AI in the OS, AI in the cloud, AI in the browser. Agentic browsers, game engines, OS/kernel-level engineering.
+
+**Tooling & Utilities** — High-performance CLI tools, networking tools, directory viewers, and system utilities.
 
 ## Why Local AI?
 
@@ -31,7 +60,7 @@ That's fast enough for interactive use. No GPU required.
 
 ## The Hardware
 
-Our target is consumer hardware. Specifically:
+Our target is consumer hardware:
 
 - **CPU**: AMD EPYC-Rome 16-core @ 2.0GHz (or equivalent)
 - **RAM**: 30GB
@@ -40,25 +69,38 @@ Our target is consumer hardware. Specifically:
 
 If it runs on this, it runs anywhere.
 
+## The Non-Negotiable
+
+All intelligence must be local AI powered. Free, local, private, offline models. No cloud. No API keys. No subscriptions. No data leaving your machine. The brain for these programs has arrived with models like Qwen 3.5 — small enough for on-device, capable enough for real agent work.
+
 ## Design Values
 
 ### Local-first
-All intelligence comes from models running on your machine. Your code never leaves your hardware. No accounts, no billing, no rate limits.
+All intelligence comes from models running on your machine. Your code never leaves your hardware.
 
 ### Zero Dependencies
-Gilgamesh uses Go stdlib only. No third-party packages. This keeps the binary small (~9.8MB), builds instantly, and eliminates supply chain risk.
+Gilgamesh uses Go stdlib only. No third-party packages. Small binaries, fast builds, zero supply chain risk.
 
 ### Lean Token Budget
-The #1 constraint for local CPU inference. We obsess over keeping system prompt and tool definition overhead minimal. ~1,600 tokens total — not 16,000.
+~1,600 tokens total overhead. Not 16,000. Every token costs latency on CPU.
 
 ### CLI + MCP + API
-Every capability is exposed through three interfaces. Humans use the CLI. Other AI agents use MCP. Programs use HTTP. Same tools, same registry, different transport.
+Every capability through three interfaces. Humans use CLI. Agents use MCP. Programs use HTTP. Same tools, same registry, different transport. There is a duality between CLI and MCP — every MCP tool works just as well as a CLI command given to an agent via bash.
 
 ### Test-Driven Development
-Gilgamesh promotes TDD in its system prompt and provides a dedicated `test` tool. It practices what it preaches with its own comprehensive test suite.
+Gilgamesh promotes TDD in its system prompt. It practices what it preaches with comprehensive tests across all packages.
 
 ### High-Performance Languages
-Go, Rust, Zig. No interpreted languages. Every millisecond matters when you're doing CPU inference.
+Go, Rust, Zig. No interpreted languages. Every millisecond matters for CPU inference.
+
+### Mythological Naming
+Projects named after gods and legends — real mythology (Gilgamesh, Zeus, Raijin, Garuda) and imagined (inspired by SWTOR's Tyth, Izax, Nahut, Scyva). Each god is a specialized, autonomous entity in the pantheon.
+
+## The Bigger Picture
+
+Gods from the Machine is more than a software project. It's art — a combination of storytelling, mythology, culture, history, and code. It draws from cyberpunk aesthetics, science fiction worldbuilding (Tron, Dune, Star Wars, Star Trek), and a technologist's conviction that software can be alive, autonomous, and god-like.
+
+Programs that go beyond the current constraints and playing fields of traditional software engineering. Literal gods in the machine.
 
 ## Open Source
 

content/roadmap/_index.md

@@ -48,9 +48,12 @@ description: "Project milestones, phases, and future plans"
   <div class="phase-dot active"></div>
   <div class="phase-title">Phase 4: Gilgamesh v0.3 &mdash; Quality & Polish <span class="phase-status" style="color: var(--cyan);">IN PROGRESS</span></div>
   <ul class="phase-items">
-    <li class="done">Go unit tests for core packages</li>
-    <li class="todo">Table-driven tests with edge cases</li>
-    <li class="todo">CI/CD with GitHub Actions (build + test)</li>
+    <li class="done">Unit tests for tools package (registry, read, glob, grep)</li>
+    <li class="done">Unit tests for MCP server (initialize, tools/list, tools/call)</li>
+    <li class="done">Unit tests for HTTP server (health, tools, error handling)</li>
+    <li class="done">Unit tests for config, context, hooks, and session packages</li>
+    <li class="done">CI/CD with GitHub Actions (build + test on push/PR)</li>
+    <li class="todo">Table-driven tests with edge cases for tools</li>
     <li class="todo">Better error messages and diagnostics</li>
     <li class="todo">Graceful shutdown for HTTP server</li>
     <li class="todo">Request timeout middleware</li>