Skip to main content

Codebase Statistics

Track the growth of the AILANG codebase over time. These metrics demonstrate the scale of AI-assisted development.

Loading codebase statistics...

What These Numbers Mean​

Lines of Code​

The AILANG project consists of multiple components:

  • Go Production Code: The compiler, bytecode VM, runtime, type system, and CLI
  • Go Test Code: Comprehensive test suites with a ~68% test-to-production ratio
  • AILANG Examples: Working examples and standard library
  • TypeScript/React: Documentation website and Collaboration Hub UI
  • Shell Scripts: Build automation, CI/CD, and development tools
  • Documentation: Design docs, guides, and reference material

Token Estimation​

The token count estimates how much context an AI model would need to "understand" the entire codebase. This is calculated at ~4 characters per token (typical for code).

Why this matters:

  • Modern AI models have context windows of 200K–1M tokens
  • The AILANG codebase (~10M tokens) requires strategic context management
  • This is why we use design docs, skills, and focused prompts

AI-Assisted Development​

AILANG is developed with Claude Code, demonstrating:

  1. Rapid iteration: 3,400+ commits across the project's lifetime
  2. Comprehensive documentation: Design-first approach with 370K+ lines of design docs
  3. High test coverage: ~200K lines of Go tests (~68% of production code)
  4. Sustainable pace: AI assistance enables consistent progress

Historical Tracking​

The timeline chart shows codebase growth across versions. Each minor "graduation" is captured at its release tag, with the latest release tracked as the current snapshot.

What drives growth:

  • v0.3.x: Core language features (types, effects, modules)
  • v0.4.x: Monomorphization and polymorphism
  • v0.5.x: AI providers, eval harness & search (M-EVAL-LOOP)
  • v0.6.x: Coordinator daemon & autonomous agents
  • v0.7.x: Observatory, chains & eval dashboard
  • v0.8.x: Cloud features & advanced coordinator
  • v0.9.x: Cloud integration & Pub/Sub messaging
  • v0.10.x: Bytecode VM & runtime
  • v0.11.x–v0.12.x: Bytecode fixes, three-tier OTEL tracing, new stdlib (gzip/tar), XLSX performance wins
  • v0.13.x–v0.14.x: Concat disambiguation, eval-suite prep
  • v0.15.x–v0.17.x: AI token-streaming helper, stdlib growth (deflate/zlib), external-consumer DX
  • v0.18.x–v0.19.x: Motoko executor adapter & eval-harness hardening, intra-package imports, FS sandbox diagnostics
  • v0.20.x: AILANG ships an LSP (ailang lsp) with VS Code install and contract verification
  • v0.21.x: 100% run_correct on VeraBench with Claude Haiku 4.5; AILANG an official VeraBench target
  • v0.22.x: Retired Gemini CLI, adopted Vertex Managed Agents API
  • v0.23.x–v0.24.x: Multi-channel notification fan-out, lenient real-world XML parsing
  • v0.25.x: Security toolchain bump (Go 1.25.11)

Methodology​

Statistics are generated by tools/generate_codebase_stats.sh:

  1. Count lines using wc -l across all source files
  2. Exclude vendor, node_modules, and build artifacts
  3. Estimate tokens at ~4 characters per token
  4. Git statistics from repository history
  5. History preserved in codebase_stats.json

The script runs automatically during website deployment, ensuring metrics stay current.

Source Code: generate_codebase_stats.sh | Data File: codebase_stats.json