Supabase CEO Jokes About Using 12 Claude Agents to "Reverse Engineer" Own Open Source Project
Supabase co-founder and CEO Paul Copplestone posted that he ran 12 Claude agents in parallel for 6 hours, successfully "reverse engineering" Supabase and providing the results in a 100% free open-source format, with a link to the supabase/supabase repository on GitHub.
The official @supabase account promptly retweeted, thanking Claude, stating that "Supabase is now fully available and open source." This repository has long been the official open-source codebase for Supabase, which has been built on open-source PostgreSQL since its inception.
Several replies pointed out that the code has always been publicly available on GitHub, and the agents rebuilt it without first checking the existing repository. Copplestone later posted similarly, claiming it took only 3 hours for Turso.
Supabase had previously launched an official Claude connector and MCP server, supporting database management through conversation, executing SQL, and deploying functions. This post was seen by the community as a humorous take on the behavior of AI agents.
Discussion and interaction focused on the state of open source and the efficiency of AI agents. True open-source projects have long been public, while the phenomenon of agents consuming resources to duplicate work has been amplified.
Source: Public Information
ABAB AI Insight
Since its establishment in 2020, Supabase has centered on open-source PostgreSQL, maintaining a public GitHub repository and aiming for a high valuation and large user base by 2026. Paul Copplestone has repeatedly emphasized open source and developer experience; this post humorously points to the already public code, shifting from product promotion to commentary on AI tool behavior.
From a capital perspective, the company attracts developers through open source and integration with AI tools (like Claude MCP), while advancing commercial hosting services. The strategic motive is to strengthen the open-source image while gaining traction from the popular AI topic, without actually changing code availability.
Similar cases involve other open-source projects being "regenerated" by AI tools from existing content. We are currently in a phase where AI agents are widely used for code generation, and users and developers are beginning to notice that agents may overlook existing open-source resources.
This essentially represents an efficiency mismatch in technological substitution. AI agents can generate code in parallel, but the mechanism lacks priority retrieval of public repositories, leading to redundant "reverse engineering" of already open-source projects, while open source itself has lowered the barriers to access.
ABAB News · Cognitive Law
The open-source repository is still there, but agents are reinventing the wheel.
Without searching first, they burn tokens.
True open source has always been there, just not checked.