Humans have always submitted crappy code. LLMs, however, do so at a much faster rate. Even the most active lousy coder is not going to be capable of submitting anything like that volume of code to multiple projects.

Humans have always been capable of social engineering and trying to sneak in malicious code. However, it's possible that as agents get better that they can do so much faster. The missing component will be compromised accounts, I think -- how many aged accounts can attackers get hold of to turn loose with agents?

Long-lived FOSS projects have tons of people who've created accounts many years ago that might be easliy compromised, but have checked out of actively participating. It's not necessarily going to throw up a red flag if a "person" shows up after a hiatus and starts contributing again.

So, there's more to it than overwhelming a single maintainer -- it's the capability to conduct a bunch of these attacks in an automated fashion if attackers can get hold of compromised accounts.

(As an aside, it's concerning that a maintainer would be pestered into accepting a questionable PR like this. I expect, though, that there are quite a few overworked people who have taken on things like Anaconda and are being measured on how quickly they close PRs.)

Odds are he’s not the first to think of this, he absolutely won’t be the last. If your agents, CI/CD pipeline, or whatever are vulnerable to this, it’s time to fix that now before something truly nasty comes down the pike.

Still, very cool, and a little jealous of the on-air staff that get to work there.

Comments URL: https://news.ycombinator.com/item?id=48054000

Points: 8

# Comments: 0

(Note: I’m not an LLM fan, don’t vibe code myself at all. But I would be unconcerned about security for the kind of things I would create if I did start doing so.)

Quite the opposite, in fact. But there’s a difference between the information being present somewhere, and a reasonable way to get that information in front of people in an actionable form.

We’re drowning in “information,” at present. But the mass media narratives that are most readily available distort things quite a bit for a lot of reasons. (Ratings, owner bias/interference, format.)

I'm not really sure what the point of encouraging new development is if the end result is "big company scoops it up and makes it shitty, but people get to enjoy it for a few brief moments before that happens."

To paraphrase a popular quote from IBM: “Executives and MBAs can never be held accountable: therefore executives and MBAs must not be allowed to make decisions.”

Slightly less flippant: The only way to stop this is to stop letting companies like MSFT gobble up smaller companies. That doesn’t seem likely in the near future, though. Once the Borg assimilate something, it’s just a matter of time before it’s digested and drained of value.

Drugs for anxiety treatment do wear off, but not the same way that weed or alcohol does: something like Celexa takes a few weeks to build up in the system, and don’t lose effect 12-24 hours later if you miss a dose. I’m not sure how long you’d have to stop before it loses efficacy entirely.

I’m not Nancy Reagan, though: I would not advise people to self-medicate with booze or pot if they’re suffering from depression or anxiety, but I’m not going to preach at anybody who is doing so and thinks it’s working for them. I will say that I’ve seen that end badly, though. I can think of three people I’m close to who’ve tried it and have had problems with addiction: all of them are now sober and (I believe) on regular antidepressants.

That sounds like a poor strategy. Expend all of your resources in one grand gesture rather than trying to push your enemy's internal factions to curtail or end the fighting?

Unlike the current US administration, Iran is playing a long game - one in which it has been isolated in many ways. Indiscriminate attacks on civilian targets is not going to win it many friends; putting pressure on the tech companies that have been buddying up to the administration and may have some sway, on the other hand, is a cheap strategy that could pay off. Iran understands that the only language that seems to matter with Trump's backers is profit; threaten that and you may have some success.

The fact that Iran has already done some damage to AWS data centers makes it seem likely they could do so again if they tried. I don't know for certain, I'm not a military intelligence expert, but the strategy of "throw the kitchen sink at it" seems like a sure loser.

I'm curious what you're basing this on, since Iran has been supplying Russia with drones, etc. for much of the war in Ukraine and so far has launched attacks into Jordan, Iraq, Kuwait, Saudi Arabia, Bahrain, Qatar, the United Arab Emirates, Oman, and Cyprus since the US began its attacks.

Iran may not be able to strike at sites in the US, but it could certainly target data centers in the Middle East with some hope of success. I'm not at all confident the current administration has accurately assessed Iran's capabilities or has the ability to protect the assets of US-based companies (or US citizens) in that region.