Claw Code: AI Reinvents Claude Code and Shakes Up GitHub
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
A Leak That Shakes the Development World
On March 31, 2026, a configuration error in an npm package accidentally exposed the complete source code of Claude Code, an AI agent from Anthropic dedicated to programming. This amounted to 512,000 lines of TypeScript spread across approximately 1,900 files that were made public. This leak was not the result of an attack or hacking, but rather a simple technical oversight.
Quickly, a link to the code circulated on X, triggering a viral frenzy. In less than 24 hours, the code was downloaded over 60,000 times. Some repositories reached tens of thousands of stars in record time, setting a new standard for virality on GitHub.
Claw Code: A Rewrite That Escapes the Laws
In this context, developer Sigrid Jin decided not to share the original code of Claude Code. Instead, she undertook to completely recreate it in Python, using AI tools to assist her. This method, known as "clean room design," involves reproducing the logic and architecture of the code without copying a single line.
The resulting project, Claw Code, quickly exploded on GitHub, accumulating tens of thousands of stars in just a few hours. While it retains the overall functionality of Claude Code, Claw Code remains legally independent, thus escaping the legal restrictions that weigh on the original code.
The Revelations of a Leak on the Value of AI Agents
The leak of the source code of Claude Code provided a unique opportunity to closely examine the workings of an advanced AI agent. It revealed that the true value of these systems lies not only in the model itself but especially in the orchestration of tools and the management of complex tasks.
Claude Code relies on a "harness engineering" logic, which involves coordinating multiple tools, managing complex tasks, and maintaining a coherent context over time. Thus, the AI drives entire workflows instead of simply generating code. This architecture confirms that AI tools are becoming execution systems, not just assistants.
As Business Insider points out, these technologies are already beginning to transform a wide range of professions, from law to medicine, by automating complex tasks. The leak shows that technological secrets reside in how they are integrated and orchestrated.
A Late Response to Rapid Reinvention
In response to this situation, Anthropic quickly acted by filing DMCA requests to remove the repositories containing the original code. GitHub followed suit by blocking the affected repositories. However, these measures came too late. Many developers had already copied, analyzed, and reinvented the code of Claude Code.
Claw Code, as a clean rewrite, completely escapes the restrictions, illustrating one of the limits of current protections. AI allows for the translation, reconstruction, and optimization of a system in just a few hours, making the defense of intellectual property much more complex.
A Turning Point for the Tech Industry
The Claw Code episode goes beyond the scope of Anthropic and shows that the innovation cycle is accelerating. A technology can be analyzed and reproduced almost instantly after its release, or even after a simple leak.
Developers, armed with AI tools, can now compete with large companies. This marks the emergence of a new way of coding, more intuitive and faster, where AI becomes an active partner in development.
Claw Code is not just a viral clone but a symptom of a profound change. An accidental leak was enough to trigger a near-instantaneous global reproduction of a proprietary system, demonstrating that technological and legal barriers are porous in the face of the speed of AI.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.