The Problem
React-era patterns exceed AI's reasoning capacity. 45% of AI-generated code has security vulnerabilities. Senior engineers are burning out on "vibe coded" spaghetti.
Forge
An opinionated toolchain for building applications with coding agents
forge new rally-hq # scaffolds a deployable app
forge feature "user auth" # AI generates from spec
forge test # verifies correctness
forge deploy # ships to productionThe Approach
Not a new framework. An opinionated workflow on top of proven tools. Build rally-hq first, extract patterns that work, then automate them.
Start Here
New to Forge? Read these in order.
Current Status
Progress
Active
What's done, what's in progress, what's next. Updated as we build.
Learnings
Living Doc
What worked, what didn't, why. The source material for the eventual doctrine.
Engineering Audit
Complete
Independent audit of research, methodology, and execution plan.
Decision Log
Append-Only
Chronological record of all decisions with context and rationale.
Six Conventions
The patterns that make AI-generated code reliable.
1
Finite Component Registry
AI selects from components, doesn't invent them
2
Server-First State
No client state management to confuse the AI
3
Typed Primitives
All functions have Zod schemas AI can reason about
4
Workflow State Machines
Explicit states, not implicit control flow
5
Property-Based Tests
Tests derived from types automatically
6
File-Based Discovery
Structure is in filesystem, not config
Research Archive
Background research from December 2025 that led to Forge. Reference material, not active docs.
The AI-Native Web Framework Thesis
Why React is wrong for agentic coding
Stack Comparison Matrix
Decision matrices for framework choices
Red Team Analysis
Layer-by-layer gap analysis of AI-native stacks
Opportunities
Four buildable opportunities identified
The Agentic Web Standard
Original specification-first approach (later revised)