Hmmm. It is how I would spread a Trojan horse. TBD.

That is amazing, he had at most just 3 hours to convert the code.

From my understanding, the code was leaked, and someone managed to convert the code and made it open source before the company can take it back and make Nintendo-style lawsuits.

But this caught my eye the most - the title:

is now FREELY AVAILABLE to Kim Jong Il

Wut??? He's not just ill, he's long dead! LOL

Code is not the core factor driving AI development. Data and processing power are

For the record Grok thinks it can tell truth from April Fools jokes.

Is Grok well equipped to evaluate whether this is an April Fools’ joke? Yes — very well equipped. I’m built by xAI to be maximally truth-seeking. For a claim like this I: Cross-check real-time web search results from credible news sources. Review timestamps and primary statements (e.g., Anthropic’s own words). Scan X for contemporaneous reactions and the original discovery post. Distinguish patterns of real incidents (packaging errors, researcher discoveries, corporate confirmations) from typical joke formats (no verification, satirical tone, post-April 1 timing, etc.). In short: real incident, not a prank. It’s a significant (if embarrassing) operational slip for Anthropic, but limited to the Claude Code tooling layer rather than the core Claude AI model.

Willing to bet it wasn’t an accident. And someone within the company got a significant chunk of change from one or more foreign sources.

Wow. Those silicon valley VCs will be annoyed.

As a one-time engineer, I know where he is coming from:

So he did what any engineer would do.

He rewrote the entire thing in Python from scratch before sunrise.

Is he connected to Flora Plios by any chance?

What do they do? A quick search tells me they are an Ai company. So why would this be significant?

And patriots can use it as well.

Gotta a wait and see how the IP laws hold up against the python copy he made

Oops.

It litteraly deletes any and all power instantly. For some reason I think there was a drop about this very thing.

Confirmed:

https://search.brave.com/search?q=Anthropic+accidentally+LEAKED+THEIR+ENTIRE+SOURCE+CODE+yesterday.+True+or+false%3F&source=desktop&summary=1&conversation=08ea9b1aa7e997fea448f3369032fe7f6065

True. On Tuesday, March 31, 2026, Anthropic accidentally published a source map file in the npm package for Claude Code (version 2.1.88), which exposed the tool's internal source code (over 512,000 lines of TypeScript across nearly 1,900 files).

Anthropic confirmed the incident was a release packaging issue caused by human error rather than a security breach, noting that no customer data or credentials were exposed. The leak included sensitive internal details such as 44 feature flags, anti-distillation mechanisms, and the architecture for unreleased autonomous agent modes like KAIROS, which were quickly mirrored across GitHub and other repositories before the package was pulled.

More:

https://search.brave.com/search?q=Where+is+Anthropic%27s+confirmation+of+the+source-code+leak+of+3%2F31%2F2026%3E&source=web&summary=1&conversation=08ea8c81bfb6d6586d532c551d58e9202969

Anthropic confirmed the leak via email statements to multiple news outlets, including VentureBeat, The Verge, CNBC, (Axios, with some interesting detail), and The Register, on March 31, 2026.

In these statements, a spokesperson (identified as Christopher Nulty in The Verge report) clarified that:

The incident involved internal source code included in a Claude Code release. No sensitive customer data or credentials were exposed.

The cause was a release packaging issue resulting from human error, not a security breach.

The company is rolling out measures to prevent recurrence.

The leak itself was discovered by security researcher Chaofan Shou (an intern at Fuzzland), who posted about the exposed source map file on X (formerly Twitter) on the morning of March 31, 2026.

where can i get the code and can i make it run a local Claude on my pc?

Claude Code is NOT THE SAME as the code for the Claude AI.

Kinda confusing, really. (And yes, I was among those confused).

Here's the difference:

https://search.brave.com/search?q=What+is+the+difference+between+Claude+Code+and+the+Claude+AI%3F&source=desktop&summary=1&conversation=08eade4dda344aedd54533e8ac192e57b599

Claude AI is a general-purpose conversational assistant accessed via web or desktop interfaces, ideal for writing, research, and brainstorming, while Claude Code is a specialized, command-line agentic tool designed specifically for software development.

The primary distinction lies in workflow and execution capabilities:

• Claude AI operates as a text-based conversational interface where you must manually copy, paste, and implement code suggestions it generates. It relies on you to provide all context within the chat window, and every message typically requires re-uploading the entire project context.

• Claude Code functions as an autonomous agent in your terminal with direct access to your file system, allowing it to read specific files, make surgical edits, run tests, execute commands, and commit changes to version control without manual intervention.

Key operational differences include:

• Context Management: Claude Code automatically scans your project directory to maintain awareness of your codebase structure and dependencies, whereas Claude AI starts fresh with each conversation and has a limited context window that hits a hard ceiling.

• Token Efficiency: Because Claude Code only reads the files necessary for a specific task rather than re-uploading your entire application, it significantly reduces token usage and costs for active development work.

• Integration: Claude AI is a chat interface, while Claude Code integrates directly into development environments (like VS Code via its terminal) to act as a development partner that can fix bugs and build features autonomously.

This is actually not a big deal, OpenAI also puts out Codex source code. Yeah it’s pretty crazy they let out all their secrets and tool usage- it’s really interesting to see how the orchestrator and memory functions work together. But, it isn’t very valuable on its own, and similar stuff is already out there.