I've found CLAUDE.md (etc.) make more sense when the agent is doing more generalized tasks rather than just coding.
If you're just having an agent produce code, then I'd agree. I'd argue that any additional context it might need ought to be included in README.md anyways, just like a human.
But I have projects where agents do more than just code, and retaining context about how I want the agents to operate, keeping track of gotchas, avoiding unnecessary steps in the future, etc., just ends up going into CLAUDE.md. It's not for me; it's for the agents. I also never edit this file directly. I may review it every once in a while to make sure there isn't anything weird in there, but I leave it to the agent to add what it thinks is necessary (or I'll tell it that something will be relevant in the future and that it should add it, etc.).
Although even this pattern isn't necessarily future-proof. Seems like skills are kind of taking that role. Still, I end up with a CLAUDE.md in these kinds of scenarios to catch all the little pieces of context that don't fit nicely anywhere else.
>Much of the angst around China's models centers on "losing the AI race". But what's the goal of this race? Is it to develop the best model? To sell the most tokens? To destroy humanity first?
Some people would say it is reaching some sort of singularity. Even if you don't buy into a more sci-fi interpretation of this, there are pretty grounded arguments one could make that there is some sort of "goal" in AI development that, if realized, would effectively make it a superweapon. Altman has been pretty vocal about his expectation that this will eventually happen and that it is his goal to be the guy to produce it. Even if it isn't some superintelligence, the ability for a machine to do something like, say, exploit cybersecurity weaknesses, is pretty worrying for entities like governments. It's the pretense used when we saw the US government ban a US model recently.
Again, you don't have to buy that "the singularity" is a real thing, but it's not hard to see that some people think some version of this is real and it is exactly what is being referred to as the goal in an "AI race."
This is an odd example to present this argument through.
I'm not defending AI bots overwhelming websites or hackers motivated by Polymarket, etc., but I don't really believe a 30 year old website with "approximately 160,000 source files serving around 2 million pages" is a good litmus test for the state of the online world. What's worse is the absurdity that this basic site offering niche data would be targeted because Polymarket depends on it for some of their bets, something that most websites don't have to deal with. Frankly, that seems like a much more interesting angle to explore than "this old website that should have been re-written multiple times over the last three decades now doesn't have a choice but to be re-written."
Oddly, it seems like the solution, at least in this specific case, is relatively simple (and ironic): pay for an LLM to re-write this site in a modern, scalable, secure way and have that LLM monitor the site to ensure it is behaving correctly. Put this thing in a modern platform behind a proper cache (i.e., throw the whole thing up in Cloudflare) and many of these problems just don't exist anymore. If this site is as basic as it seems, this could be a weekend project.
It's certainly an interesting story, and I don't blame the owner of The Numbers for handling his site the way he has, but framing this as "AI is destroying our beloved internet" just seems obtuse, at least through this lens.
>but may make it difficult for retail investors to recognize risks
Ok, so just to be clear: institutional investors are (a) the ones investing the large proportion of capital in these companies and (b) are well equipped to decipher financial statements. The idea that any significant amount of retail investors have even seen a financial statement, let alone is making decisions based on their analysis of a financial statement, is laughable. And, even then, if someone is putting in that effort, then presumably they're not going to get tripped up by a legitimate practice that they ought to specifically be looking for given the context.
And all of that doesn't even take into account that every article discussing the financials of AI firms over the last half a decade have been pointing out these dynamics. We're literally in a thread discussing this exact dynamic. Retail investors are certainly far more likely to make investing decisions based on these kinds of articles and threads than they are based solely on independent financial statement analysis that they're conducting. At a minimum, I think anybody taking any of this seriously has gotten the hint by now.
If it comes out that these firms are committing straight-up fraud, then there will be a lot more to discuss. But, as of now, the sentiment is that these firms are behaving perfectly legitimately, just abnormally and maybe irresponsibly compared to their historical context. If an investor isn't equipped to handle this kind of analysis under these circumstances, then I'm not going to feel too bad if they lose their money "investing" when they're really just gambling.
If you're just having an agent produce code, then I'd agree. I'd argue that any additional context it might need ought to be included in README.md anyways, just like a human.
But I have projects where agents do more than just code, and retaining context about how I want the agents to operate, keeping track of gotchas, avoiding unnecessary steps in the future, etc., just ends up going into CLAUDE.md. It's not for me; it's for the agents. I also never edit this file directly. I may review it every once in a while to make sure there isn't anything weird in there, but I leave it to the agent to add what it thinks is necessary (or I'll tell it that something will be relevant in the future and that it should add it, etc.).
Although even this pattern isn't necessarily future-proof. Seems like skills are kind of taking that role. Still, I end up with a CLAUDE.md in these kinds of scenarios to catch all the little pieces of context that don't fit nicely anywhere else.