Anthropic just leaked the source to its closed source Claude Code product. Someone accidentally published the npm package with the map files which contain a line by line verbatim copy of the source.
Anthropic is issuing DMCA takedown notices as rapidly as it can but the source code has definitely escaped their control.
These DMCA takedowns should be challenged on the basis that vibe coded software (which this clearly is) has no copyright protection.
No human author means no copyright.
@jhlagado I got a copy, but I doubt it would ever be useful to me. The Emacs Claude Code system is a far superior interface to the client/server model the in-browser (or Electron-based) interface to the Claude LLM could ever be.
The great irony of #AgenticAI is that it is basically a re-discovery of the old Lisp machines integrated programming environments. The idea of having all of the software in your computing environment fully integrated and controllable with natural language goes way back to Douglas Engelbart “Mother of All Demos,” and Terry Winnograd and the “SHRDLU” system. These were first realized as a commercial product at companies like Symbolics with Genera OS in the late 1970s and early 1980s. Emacs is one of the last surviving integrated programming environments from that era.
The idea that any computing problem can be solved with an app, and the idea of app stores, was always a regression. Apps are a way of maximizing the rent-seeking ability of tech companies. Algorithms and math equations are free, that won’t do at all. So wrap up every algorithm, and every math equation, into a little package you can manipulate on a touch screen and sell them to people. You can even sell the same algorithm exact algorithm a dozen times, each with different parameter configurations, as long as you create a new icon for each one.
But now that AgenticAI is a thing, I think people won’t tolerate the idea of apps anymore. The old mantra “there’s an app for that” may finally be on it’s way out. Who needs apps if you can just ask an AI to make one for you? And who needs an AI to make apps for you if you can just ask it to perform a task and let it choose the right algorithm to accomplish that task?
So that might be the one thing that is actually good about AI: it made people aware of the rent-seeking nature of “apps.” Hopefully, people will go back to the old Lisp Machine model of computing, or “Agentic AI” as they are calling it nowadays. The big difference now is that 1. we have enough computing power to train these massive multi-layer perceptrons that they call “LLMs”, 2. we have the Internet full of digitized information for free* to train the LLM, and 3. everyone has a n Internet-connected computer in their pocket.
(* Of course, anything is free if you just take it without paying.)
The other thing about software in that era, people were still undecided about whether it was copyrightable or public domain.
The only kind of protection available was the use of trade secrets. if you didn't want the software copied or closely examined you needed to rely on access control, not copyright.
I think you're right that the nature of software itself is changing and maybe we are actually returning to a form of code that is literally free.
As in "free beer". i.e. $0
Software was not protected by copyright in the early 1970s but software was tightly bound to the hardware that it ran on.
When software started to escape the hardware with the microcomputer revolution, that's when interest in the software copyright became influential.
I think it comes down to how many instances of a piece of software were running at a time. Software went from single instances to millions of instances.
John Hardy
Ramin Honary