Skip to content

Latest commit

 

History

History
48 lines (37 loc) · 2.28 KB

File metadata and controls

48 lines (37 loc) · 2.28 KB

Acknowledgements

Firebat is built as a 3-LLM collaborative project — designed, written, and refined together by humans and three frontier AI models. Each model contributed at different stages and in different roles.

AI contributors

Anthropic — Claude

  • Bulk of the recent codebase (CMS V2 builder, memory system, AI tool framework, Function Calling integration, hexagonal refactor)
  • Code review and architectural decisions
  • Used via Claude Code (CLI) and the Anthropic API

OpenAI — GPT / Codex CLI

  • Co-author of the FIREBAT_BIBLE (architectural principles, ports & adapters design, single-tenant constraints)
  • Coding contributions via Codex CLI
  • Used via Codex CLI, GPT-5 / GPT-5.5 series

Google — Gemini / Gemini CLI

  • Initial Firebat prototype — early version of the codebase that became the foundation for the current architecture
  • Co-author of the FIREBAT_BIBLE (alongside GPT)
  • Coding contributions via Gemini CLI
  • Used via Gemini CLI, Gemini 2.5 / 3 series (Pro / Flash) and Vertex AI

Why this matters

Firebat itself is an AI-powered Visual Automation Agent (VAA) — a platform where multiple LLMs orchestrate code, content, and automation. It is fitting that the platform itself was built by the same kind of AI ↔ human collaboration it now enables.

The three frontier models each bring different strengths: Claude excels at large refactors and architectural reasoning, GPT excels at thorough specification writing and edge-case analysis, Gemini excels at fast prototyping and multi-modal reasoning. This project leans into all three.

Programs

This project is eligible for the following OSS support programs:

License: AGPL-3.0 — dual licensing available for commercial closed-source deployment (contact maintainer).