stingraycharles

Yes but my point is that I never realized that the AI would be writing tools to update the weights directly, rather than improve the existing training process.

A tangent, but I’m now realizing that the “recursive self improvement” thing that these LLM companies are talking about is (probably?) not about unsupervised Claude Code instances improving the training process, but rather LLMs modifying their own weights through tooling directly. It would certainly open up a whole new process for model optimization.

Yeah ok of which $1.1B is stock options, so more like $40k a month in actual salary.

I can only assume the whole announcement is about enabling http compression.

I'm amazed that they're not ashamed that they weren't compressing their requests all this time.

Aren’t the concepts mostly already in place for a decade or two at Meta and Google?

Which would support your argument, but I don’t think this is particularly new.

Assuming it’s entirely unsupervised, yes, which I can only assume to be the case at these volumes (hence the dogfooding).

I do believe this will force the AI labs to eventually solve the bloat problem with AI, but we’re definitely not there yet.

I’m not buying this argument, the apps of the LLM providers are some of the most wasteful apps that exist. Have you ever looked at how much memory / CPU Claude and Codex use? They ship with entire virtual machine hypervisors for sandboxing.

I think it’s simply a matter of dogfooding and new feature development, they do have much higher velocity than companies of equivalent size and an unlimited token budget helps as well.

Not in that way, more that they can predict what good or bad relations between weights would look like.

And what’s the math? OpenAI is not going public, Anthropic is, and the math is… the SEC will block the Anthropic IPO because of what exactly?

It’s astonishing how many people here claim to be financial experts and interpret the Anthropic IPO in a way that makes them look evil.

That’s the old style IPO though, most IPOs nowadays do not actually raise money.

SpaceX made that even more clear by doing an explicit offering of shares to raise money just months after their IPO, which didn’t raise them any money at all but was just for investors and employees.

To let early investors exit? Most private funds like that have a horizon of 10 years, so they want to sell their investments at some point.

If the SEC allowed the SpaceX IPO they most certainly will allow the Anthropic IPO. From my point of view, the Anthropic IPO is a much more genuine one than the shenanigans that were going on with SpaceX, yet we’re all investing into that through ETFs as well.

Additionally, some ETFs have built in protections to rule out businesses that are not yet profitable (like SpaceX and Anthropic), which I think is a great rule.

Well what I do know as someone who has a bird’s eye view into many enterprises is that many of them are starting to realize they’re bleeding money to Anthropic and are telling their teams to be more efficient with tokens.

Don’t forget that weights of models and their behavior may, perhaps unintentionally, be in their training set.

Or at least those of some models, and I wouldn’t be surprised that an LLM model is able to understand correlations between weights and behavior better than humans could ever do.

There are two types of multipath TCP and they are not at all the same.

What Apple does is, for example, being able to have a connection over WiFi and 5G simultaneously, with different IPs and routes.

The more traditional server side multipath is that you have a network card with 4 ports (eg 4x 25gbit) and connect them to a router, and make those 4 physical ports (paths) behave like a single interface. This helps either load balancing or redundancy.

The latter is what this vulnerability is about.

19 hours in and it's still the same, I don't understand why it's not being updated

Wow if that is the case… Cursor certainly needs to communicate much more clearly that Anthropic wasn’t involved in this at all.

I’d still expect a post mortem, be it from Cursor or Anthropic.

This assumes that it’s more profitable for these labs to compete with their own customers, rather than increasing the productivity for their customers and - assuming this increase justifies it - increasing their profits.

I don’t think it will ever reach such a point.

It’s strictly speaking not their responsibility, but it is something that they should do if they want to help their customers.

In general, they need to make a really big post somewhere that’s not a HN comment that this was in fact a compromise of the Cursor marketplace, not an employee’s machine, because they’re getting a lot of (unjustified) bad PR.

I think you’re being extremely negative towards Anthropic here. Basically the parent is saying “this is a problem at Cursor and not Anthropic” and you’re holding them accountable.

You’re literally saying that they should not have published on the Cursor marketplace but should have used OpenVSX because that would have been “reasonable”, but I think you’re being unreasonable here.

It means that Anthropic did not intentionally develop a malicious version, obviously. But it’s a bit disingenuous and tries to suggest that it wasn’t Anthropic that uploaded it and/or wasn’t published under Anthropic’s name, which isn’t true.

I’d expect nothing less than a post mortem on how this happened to be honest.

And try to launch sshd, right? Think of all the other things they could have done…

For what it’s worth, at one point one of our employees windows machines got infected, and the moment he did $something with Wordpress it intercepted the HTTP post and injected malware in there.

So that’s how these things can go, the employee’s machine may have been infected which was then used to inject this.

Wasn’t Elon Musk shouting a few days ago that neoclouds are going to crash, the money dried up and that SpaceX was the last company with the opportunity to raise more money?

I think you’re giving Musk too much credit.

I use Claude Code extensively as I mentioned, but what does “primary support” look like? Why should tmux “know” what an agent is?

I just use tmux with multiple ghostty sessions and I have different Claude code sessions behind different tabs and panes and work like that.

What am I missing out on that justifies now using a normal terminal+tmux anymore?

But I’m already using ghostty. I installed herdr but made the mistake of launching it from within tmux and had no idea how to close it.

I then opened it somewhere else, but I had no idea how to interact with it. I also couldn’t get two other windows to “join” the same herdr session, so I gave up and went back to tmux.

For someone who has used screen / tmux for almost 30 years at this point, what does herdr do that tmux can’t?

I use Claude Code extensively. I also use plannotator extensively, which is how I found out about it.