AI wobbles, leaks – now what?
are AI large laguage models ighting and hacking each other?
The AI wars seem to have commenced with a leak of source code from ‘Anthropic’s large language model - ‘Claude’.
From here; (h/t CITIZEN FREE PRESS)
You will have to read the threads in the twitter feeds ad try and make sense of the ‘leak’ ad ‘afer all his time. “ what difference does it make?”.
‘This is partly true, but heavily exaggerated.’
‘What appears well supported is that Anthropic accidentally exposed a large chunk of Claude Code’s source through a bundled source-map file in npm package version 2.1.88; multiple outlets reported roughly 512,000 lines across about 1,900 files, and Anthropic told Decrypt it was “a release packaging issue caused by human error,” not a breach.
It also appears true that Chaofan Shou was among the first to spot it and post a download link on X, and that mirrors and forks spread quickly afterward.’
‘Mirrors and forks’ at 4 am. OK!
‘The middle of the story is directionally plausible but not fully nailed down.
Reports say Anthropic pulled the package and that mirrors were hit with DMCA takedowns, while a repo called claw-code by Sigrid Jin / instructkr does exist, presents itself as a Python clean-room rewrite, cites the Wall Street Journal’s note that Jin used 25 billion Claude Code tokens last year, and says a Rust port is in progress.
The GitHub page currently shows about 51.7k stars and 56.6k forks, so the “49,000 stars and 56,000 forks” claim is now in the ballpark, though I did not find a primary source proving it was the fastest repo in GitHub history to hit 30,000 stars.
That “history” claim is repeated by the repo’s own README and by Decrypt, but I would treat it as unverified marketing unless GitHub itself confirms it.
The weakest parts are the more cinematic details.
I found support for 16 million views on Chaofan Shou’s thread from Decrypt, not 21 million.
I also found reporting and summaries around “Undercover Mode” in the leaked code, but not a primary Anthropic statement confirming the broader dramatic framing around it.
And “Anthropic leaked their entire source code” is overstated: the leak appears to involve Claude Code’s CLI/harness source, not Anthropic’s model weights or the company’s whole codebase.
So the clean verdict is: real leak, real scramble, real mirrors, real rewrite project — but the viral post inflates several numbers and overstates the scope.
From Jeremy’s thread;
‘Anthropic pushed a software update for Claude Code at 4AM.
> A debugging file was accidentally bundled inside it.
> That file contained 512,000 lines of their proprietary source code. > A researcher named Chaofan Shou spotted it within minutes and posted the download link on X.
> 21 million people have seen the thread.
> The entire codebase was downloaded, copied and mirrored across GitHub before Anthropic’s team had even woken up.
> Anthropic pulled the package and started firing DMCA takedowns at every repo hosting it. > That’s when a Korean developer named Sigrid Jin woke up at 4AM to his phone blowing up.
> He is the most active Claude Code user in the world with the Wall Street Journal reporting he personally used 25 billion tokens last year.
> His girlfriend was worried he’d get sued just for having the code on his machine.
> So he did what any engineer would do.
> He rewrote the entire thing in Python from scratch before sunrise.
> Called it claw-code and Pushed it to GitHub.
> A Python rewrite is a new creative work. DMCA can’t touch it.
> The repo hit 30,000 stars faster than any repository in GitHub history. > He wasn’t satisfied. He started rewriting it again in Rust. >
It now has 49,000 stars and 56,000 forks.
> Someone mirrored the original to a decentralised platform with one message, “will never be taken down.”
> The code is now permanent. Anthropic cannot get it back. Anthropic built a system called Undercover Mode specifically to stop Claude from leaking internal secrets. Then they leaked their own source code themselves. You cannot make this up.
You need to be a expert in DMCA, forks and mirrors to understand any of this and its implications for AI in general.
I am not!!
I have no idea whether this increases or decreases the demand for larger or smaller data centers.
I use Brave AI a lot and it is not mentioned, but maybe it is a ‘lightweight’ in this arena.
Please take a (paid or unpaid) subscription or forward this article to those you think might be interested.
You can also donate via Ko-fi – any amount from three dollars upwards. Ko-fi donations here:
Did anyone spot the code area in which Claude participated in Project Maven to target military sites in Iran, including the girls school where 170 innocent souls were slaughtered by these AI demons?