Anthropic is removing its covert code for catching Chinese competitors Anthropic says that it plans to remove hidden codes it added to Claude Code several months ago to catch other AI companies that are trying to steal from its models. Thariq Shihipar, an engineer at Anthropic who works on the Claude Code team, said on Tuesday that a fix should appear on July 1. "This is an experiment we launched …
So, you were upset about other folks (maybe Foreigners!) scraping the content that you'd scraped from folk around the world? We had to put in mitigations because of AI bots, mostly Claude, killing the servers on our small, UK, forum. 40,000 sessions in any 8 hour period, compared with less than 100 genuine users...
>>The cowboys at Anthropic and OpenAI cannot even write a decent scraper. I guess it is because they are all fresh out of uni research departments and so have no real world experience.
No, it's because their tools didn't scrape any decent scraper code and were left trying to write it themselves. Poorly.
If there's one thing that makes me think AI is a bubble, it's the lousy quality of the code in the leaked Claude Code source and now this leak.
If their customer-facing software is so hacked-together and amateurish, then how bad is it in the billion dollar back-end? Do these high-value AI companies actually know what they're doing?
A purely rhetorical question I take it.
The answer is patently clear to all except the companies themselves although a case might be made for the likes of Altman knowing exactly what they are doing which is a far older practice than that of AI. I imagine Bernie Madoff might be glad to have a few new faces on his block.
Leaked source code is always of poor quality.
Those of us who write high-quality code are not ashamed to let the whole world see exactly how good it is.
The wide-scale adoption of a single machine architecture, and the flood of proprietary, caged software that it enabled to be supplied in binary-only form, was a gift to the world's worst coders.
Chinese and other open weight models are the biggest threat to a business model that is close to charging people for thinking. This is economic nonsense and somewhat akin to having only very expensive private education that few can afford. Knowledge is best considered a public good.
At least when it comes to programming, the open models are now considered only a couple months behind the best commercial ones with the gap closing all the time. They tend to be more resource intensive and less optimised but those are the kind of features that healthy competition will deal with, especially are more users switch to them.
I always used to have a deap-seated fear that such plagiarism traps might have been hidden somewhere in a book of mathematical tables, and an engineer relying on it for an innocent calculation -- using the original, mind you, not a "pirate" copy -- would accidentally hit upon one of them; ultimately resulting in something like the collapse of a bridge, bursting of a hydro-electric dam or an aircraft exploding over a populated area.
