71–80 of 817 posts
I used Claude Code for a month because my boss gifted me a sub and wanted me to try it. I used that month to complete a work project and then beef up my personal harness so I'd never have to deal with Anthropic (and these sorts of shenanigans) again.
[flagged]
This is very interesting. Combating resellers and distillation seems like a very difficult problem indeed. Interesting to me is that these techniques mentioned in the article are just like anti-observation techniques used by some of the more sophisticated malware out there, however defeating them is pretty trivial.
Yes, defeating this is relatively easy, particularly for sophisticated actors. But it's hard to always defeat all of the tricks. Sort of like how it's expensive and hard and uncertain to defeat all of the tricks when forging money. Here's an example. Say you have your team use patched binaries. Then CC updates and requires a new patched binary with new tricks. You now have to have a team ready to analyze the binary a…
Cool reverse engineering/analysis report but if this is the extent of nefarious activity that came of it (trying to catch/mitigate chinese lab model distillations), that's kind of encouraging.
If they only collect the data for analysis I guess this is fine (they already get way more sensitive data from users anyways, so if privacy is your concern you've made the mistake many steps ago). The much more interesting question is if they directly act on this data in their API. For example by rate-limiting, compute-limiting or rerouting to weaker models. That might even be legally questionable. I would really lik…
I used Claude Code for a month because my boss gifted me a sub and wanted me to try it. I used that month to complete a work project and then beef up my personal harness so I'd never have to deal with Anthropic (and these sorts of shenanigans) again.
How do people build something like a personal harness? Are there tools for that or is it done from scratch?
Me, personally, I didn’t build it from scratch but I ported original CC from published sources into Python and extended it to match my own requirements.
Earlier quoted context omitted.
Yes, defeating this is relatively easy, particularly for sophisticated actors. But it's hard to always defeat all of the tricks. Sort of like how it's expensive and hard and uncertain to defeat all of the tricks when forging money. Here's an example. Say you have your team use patched binaries. Then CC updates and requires a new patched binary with new tricks. You now have to have a team ready to analyze the binary a…
Is it hard? Just ask AI if the update added any new fingerprinting vectors?
Earlier quoted context omitted.
> Would you also say that "someone who wants to use an IDE / LSP features to code and not give credit to the IDE / LSP is the worst kind of person"? That's a false equivalency. > If not, what is the difference between the two for you? Let's start this out right: if they're equivalent, first you explain to us why you think so.
> That's a false equivalency. How is it false? > Let's start this out right: if they're equivalent, first you explain to us why you think so. I think it should be really obvious how they're equivalent: both are the result of a program running on a computer, and not the result of in-the-moment cognition by a moral agent or moral patient. Of course the LLM is just a tool. Models can literally be downloaded as ordinary…
In fact it's really obvious everything is equivalent: it's all just matter and energy!
> Of course the LLM is just a tool. Models can literally be downloaded as ordinary files. There is not some threshold to cross where some configurations of bits on a disk deserve "credit" for work and others do not.
Of course there is such a threshold. And it's definitely been crossed when the "tool" can operate autonomously or nearly so, when it can generate the "creation" with minimal operator input or understanding.
Your classic IDE can't do anything without the detailed control of its operator. It's nothing like a coding agent.
Can somebody clarify for me - if ANTHROPIC_BASE_URL is set to a different provider... then isn't this "marked" system prompt being sent to that provider's API rather than Anthropic's? I understand how this can be useful to Anthropic if the 3rd-party is acting as a proxy (because they end up hitting the Claude API with the marked prompt), but it looks like requests where "hostname contains deepseek" would never be sen…
https://www.chinatalk.media/p/how-to-buy-cheap-claude-tokens...