This is what I'd expect from companies - I don't see why Facebook would get money because they helped people connect to each other who ended up developing a cancer cure, but they definitely should be held accountable for enabling a genocide. You're allowed to operate a business until you cause harm to society, then we can shut it down.

I think the big thing you would need is to see the internal emails - if there was ever a case where someone raised a concern about this possibility and it wasn't taken seriously, then they should be liable. If they just never thought about it then it could be negligence but I think if I was on a jury I'd find that more reasonable than knowing it could be a problem and deciding you aren't responsible

> I don't see why Facebook would get money because they helped people connect to each other who ended up developing a cancer cure, but they definitely should be held accountable for enabling a genocide.

Why? What does it even mean to "enable a genocide"? Just saying something isn't an argument.

> if there was ever a case where someone raised a concern about this possibility and it wasn't taken seriously, then they should be liable.

Again, why? How is this any different than electricity as a tool, which has both beneficial and harmful uses? AI is knowledge as a utility, that's the position here.

We've also seen this within the US - California generally makes the first move and then companies just follow that law because they know others might change and it's easier than doing it by state. One relatively small law can have a big impact, we also follow GDPR in the US because a lot of companies operate in europe too

A lot of this can be provided or built up by better documentation in the codebase, or functional requirements that can also be created, reviewed, and then used for additional context. In our current codebase it's definitely an issue to get an AI "onboarded", but I've seen a lot less hand-holding needed in projects where you have the AI building from the beginning and leaving notes for itself to read later

Curious to hear if you've seen this work with 100k+ LoC codebases (i.e. what you could expect at a job). I've had some good experiences with high autonomy agents in smaller codebases and simpler systems but the coherency starts to fizzle out when the system gets complicated enough that thinking it through is the hard part as opposed to hammering out the code.

I'd estimate we're near a million LoC (will double check tomorrow, but wouldn't be surprised if it was over that to be honest). Huge monorepo, ~1500 engineers, all sorts of bespoke/custom tooling integrated, fullstack (including embedded code), a mix of languages (predominantly Java & JS/TS though).

In my case the AI is actively detrimental unless I hand hold it with every single file it should look into, lest it dive into weird ancient parts of the codebase that bear no relevance to the task at hand. Letting the latest and "greatest" agents loose is just a recipe for frustration and disaster despite lots of smart people trying their hardest to make these infernal tools be of any use at all. The best I've gotten out of it was some light Vue refactoring, but even then despite AGENTS.md, RULES.md and all the other voodoo people say you should do it's a crapshoot.

Ask the AI to figure out your code base (or self-contained portions of it, as applicable) and document its findings. Then correct and repeat. Over time, you end up with a scaffold in the form of internal documentation that will guide both humans and AIs in making more productive edits.

If you vector index your code base, agents can explore it without loading it into context. This is what Cursor and Roo and Kiro and probably others do. Claude Code uses string searches.

What helps is also getting it to generate a docs of your code so that it has map.

This is actually how humans understand a large code base too. We don’t hold a large code base in memory — we navigate it through docs and sampling bits of code.

Around 250k here. The AI does an excellent job finding its way around, fixing complex bugs (and doing it correctly), doing intensive refactors and implementing new features using existing patterns.

Our codebase is well over 250k and we have a hierarchy of notes for the modules so we read as much as we need for the job with a base memory that explains how the notes work

cloc says ours is ~350k LoC and agents are able to implement whole features from well designed requirement docs. But we've been investing in making our code more AI friendly, and things like Devin creating and using DeepWiki helps a lot too.

Since when do we measure productivity by lines of code?

It’s not a measure of productivity, but some number of new lines is generally necessary for new functionality. And in my experience AI tends to produce more lines of code than a decent human for similar functionality. So I’d be very shocked if an agent completing a feature didn’t crank out 1500 lines or more.

We have this in some of our projects too but I always wonder how long it's going to take until it just fails. Nobody reads all those memory files for accuracy. And knowing what kind of BS the AI spews regularly in day to day use I bet this simply doesn't scale.

Not blind, and can't speak to how popular or useful they are, but there are products meant to be used like that [0]. I can't find the link but I've also seen this done with paintings, where someone creates essentially a sculpture based on a painting, and then they can 3D print it so a blind person could "see" something like the Mona Lisa or Starry Night.

A while ago I read a biography of Louis Braille, and he created his system to replace an older one where they would teach people to feel the shape of letters in wooden blocks. Braille replaced it because it was much easier to read fast, but it was never meant to be used for something like a picture.

I'd also be interested if something like a tactile floor plan would even be useful for someone blind from birth, from what I've heard you don't think about navigating spaces the same way, so a floor plan might be far away from the mental models they use.

[0]: https://evengrounds.com/services/tactile-3d-printed-models-f...

Sometimes I draw UML-like diagrams when I join a project (and when the project is big enough in such a way my mind melts if I try to keep track of everything), I wonder if there are equivalent representations of such things.

Linear text is perfect to me for documentation, teaching/learning etc...

But also, systems seems to be better digested under the shape of spatial representations (I met a lot of CS persons that fantasized over the possibility of displaying all the files of your codebase in a VR-like environment augmented with visual cues (UML) and I must admit that I would find this unnecessary but comfortable -- and I can imagine applications of this in other domains; imagine a teacher arranging the whole of Kant philosophy as a spatial-visual system referencing positions, comments, etc..). Eyes are cool because you can focus on something while knowing that some available information is there around the zone you are focusing; in a sense, so is the hand, locally, but I imagine (I dont know) it would require some super-human level of braille reading to be able to jump back and forth reading on different fingers, so that's again a probably stupid question to ask to the blind crowd of hn : are you able to do this?

evengrounds.com mentioned recently: https://news.ycombinator.com/item?id=46502269#46547839

Anecdotally, I have used SO much less in recent years, but I think a lot of what gets labelled as toxicity is just dedicated curation. I remember asking my first question - I got a quick response that it wasn't well written and was likely a duplicate, and they were totally correct.

The next time I asked I made sure to put the work in beforehand to make sure it was well written and included any relevant info that someone would need, as well as linking potential duplicates and explaining the differences I had to them. That got a much better response and is more useful for future readers.

I'm sure this sort of unofficial blacklisting is fairly common, but it does seem very opposed to the idea of a free market. It definitely doesn't seem like Anthropic was trying to make some sort of point here, but it would be cool if all the AI companies had a ToS saying it can't be used for any sort of defense/police/military purposes

I am not even sure what free market is, aside from Economics textbooks and foreign policy positioning. Whatever it may be, I don't think we had it for quite some time.

> It’s nuts that personal computers aren’t personal anymore

I think the core driver here is that most people don't want a "personal" computer, they want a device that's able to reliably accomplish tasks. Early computers gave users much more power and control but that also came with the responsibility to set up and maintain the system, which limited the userbase a lot. I'd argue a lot of the security is security from the user against themself - there is definitely some value in trying your best to make sure a user is unable to brick their phone no matter how much they try, because they're likely going to blame you and ask you to fix it afterwards

We can have both worlds.

Just because there are some users that prefer such devices doesn't mean that a technical user who can repurpose that device shouldn't be allowed to.

I know it's not quite the point of the article, but just to push back on the phrase

> I should be able to run whatever code I want on hardware I own

There's a few cases where this definitely seems wrong - you can own a radio transmitter but it's super illegal to broadcast in certain frequencies. So while you're "able" to in the sense that's in physically possible, you're not "able" to because it's illegal, and I think most people would want it that way.

In a similar way, it's illegal to modify your car or especially guns in certain ways. I could see a similar argument saying "I own this machine, I should be able to modify it mechanically however I want". Yes you own it, but as soon as you bring it in the world then you also need to account for how it's going to impact everyone else. You can't even manufacture certain hardware on your own without the right approval.

If it's "I should be able to run whatever code I want on hardware I own if I accept the risks of doing so" then that seems more balanced, but also doesn't seem too desirable because you're adding more footguns into the world that average consumers wouldn't want to run into accidentally

OP here: Yes I agree that there needs to be limits on this. Classic "your freedom ends where mine begins"

OP does address why it's critical:

> Not that we need to re-litigate that homelessness as a national security issue

Without this it's easy to think that this was just a bad actor we could have caught, instead of just a symptom of a deeper issue not being addressed

I'd be more surprised if there isn't a causal link between homelessness and making bad choices - I don't think it's really disputed that there's a causal link between homelessness and crime in general.

Amusingly, a significant fraction of people will read what you wrote as a causal link in a particular direction and agree with it. And a different fraction of people will read it as the link going in the other direction and also agree.

I'd hope that it's not a significant fraction that would get it backwards haha but just to clarify - homelessness/poverty will cause people (on average) to make worse decisions and lower cognitive ability, but making bad decisions or having a lower cognitive ability is not a cause of homelessness/poverty, at least from a statistical causality perspective on a population, individual cases will of course be different

> being hard with people that just say hello just doesn't make much sense to me anymore

This is the problem I run into, I want to just reply to any "Hey" message with a link to this page, but then I'm the one being rude. We just need a better way to let other people know that this isn't a good way to do async chat. I've heard of other people making their status message this site, so then people see it when they go to message you and it doesn't have to be explicitly brought up

> If the worst they do to me is to say hello and never talk to me again, I'm ok with accommodating this in my daily workflow

This I can't really get behind, because if they just send a hello it's implied that I then need to follow-up and find out what they were asking about