331–340 of 397 posts
Earlier quoted context omitted.
But LLMs will get that code for the next raining and planty of people will use it productive, just look how many people use OpenClaw. There are people for whom a software that compiles without error is for productive use cases
> Never make toy software and share it! > Someone might try to use it and get pwned!
The future is that people stop buying software and just build it themselves. The spam filter in thunderbird was broken for me, I built my own in hours and it works way better. Oh that CRM doesn’t have the features you want? Build one that does. It will become very easy to built and deploy solutions to many of your own bespoke problems.
Had an experience like this recently. QEMU stopped compiling for old versions of MacOS (pre-13) w/M1 arch, due to it requiring newer SDKs which don't support older MacOS versions. I put Sonnet 4.6 on the case, and it wrote a small patch, compiled and installed it in a matter of minutes, without giving it any instructions other than to look at errors and apply a fix. I definitely would have just given up without the A…
Earlier quoted context omitted.
He also mentioned it took 2 months. I’m actually wondering how long it would take to do the Linux to BSD port by eyeball, or at least ai assisted. Probably not that much longer? I guess it depends on wall time vs real time.
I estimate two weeks from having never seen kernel source to something reasonably stable based on experience with block devices/raid controllers. But I knew a bit of C (had patches merged into SVN, Exim4, etc).
Earlier quoted context omitted.
>anything wrong with that? Wonder if the courts will move fast enough to generally matter.
As I understand it, reverse engineering for the purpose of interoperability is allowed under the law. The only thing subject to copyright is your code. So long as a separate implementation (made by an AI model or made by hand) doesn't use any of your actual code, you have no claim over it. Only the code is yours. AI models make the process of reversing and reimplementing drivers much cheaper. I don't understand the p…
Earlier quoted context omitted.
> Nothing to do with AI, or even the capabilities of AI. The person intentionally didn't put in much effort. The part to do with AI is that it was not able to drive a comprehensive and bug free driver with minimal effort from the human. That is the point.
Why is that the metric? In my job, I get drafts from junior employees that requires major revisions, often rewriting significant parts. It’s still faster to have someone take the first pass. Why can’t AI coding be used the same way? Especially if AIs are capable of following your own style and design choices, as well as testing code against a test suite, why isn’t it easier to start from a kind of working baseline th…
Had an experience like this recently. QEMU stopped compiling for old versions of MacOS (pre-13) w/M1 arch, due to it requiring newer SDKs which don't support older MacOS versions. I put Sonnet 4.6 on the case, and it wrote a small patch, compiled and installed it in a matter of minutes, without giving it any instructions other than to look at errors and apply a fix. I definitely would have just given up without the A…
Why would you solve an issue like this and then not supply a patch upstream, or at the very least contact someone that could? It seems to be like the FLOSS equivalent of posting about a problem on a forum and then replying "nvm, solved it".
Had an experience like this recently. QEMU stopped compiling for old versions of MacOS (pre-13) w/M1 arch, due to it requiring newer SDKs which don't support older MacOS versions. I put Sonnet 4.6 on the case, and it wrote a small patch, compiled and installed it in a matter of minutes, without giving it any instructions other than to look at errors and apply a fix. I definitely would have just given up without the A…
Why would you solve an issue like this and then not supply a patch upstream, or at the very least contact someone that could? It seems to be like the FLOSS equivalent of posting about a problem on a forum and then replying "nvm, solved it".
Had an experience like this recently. QEMU stopped compiling for old versions of MacOS (pre-13) w/M1 arch, due to it requiring newer SDKs which don't support older MacOS versions. I put Sonnet 4.6 on the case, and it wrote a small patch, compiled and installed it in a matter of minutes, without giving it any instructions other than to look at errors and apply a fix. I definitely would have just given up without the A…
Why would you solve an issue like this and then not supply a patch upstream, or at the very least contact someone that could? It seems to be like the FLOSS equivalent of posting about a problem on a forum and then replying "nvm, solved it".
With some effort OP could review it manually and then try to submit it though.
But QEMU uses a mailing list for development, it's tedious to set up and then later keep track of. I now fundamentally refuse to contribute to projects that use mailing lists for development, the effort it takes and the experience is just so horrible.
Especially if it's a small patch that doesn't concern anyone (any big sponsors), you'll probably never get a response. Things get lost easily.
Earlier quoted context omitted.
Funnily enough, this will make many "tragedy of the commons" / "Goodhart's law hacking" problems more tractable. Right now, there's only one Google algorithm, one Amazon search and so on. The moment you let agents run wild, each with a different model, prompt and memory, effectively introducing randomness into the process, it becomes much harder to optimize for "metric go up."
we've seen what "no barrier of entry" marketplaces look like... Quality go down.
The quality will always be lower for a new product/ production line, because 1) it hasn't had the time to iterate that got the established, big-name producers to where they are, and 2) it democratizes the market to allow for lower-quality version that weren't fiscally feasible under a more complex (and thus expensive) manufacturing/ production base.
But after the market normalizes, it will start to naturally weed out the price-divorced low-quality products, as people will figure out which ones are shitty even for their price, and the good-for-their-price ones will remain.
Eventually you'll end up with a wider range of quality products than you started with, at a wider range (especially at the low end, making products more accessible) than when it started.
High barrier of entry marketplaces only benefit big companies who don't want to actually compete to stay on top.
Tying it back to the discussion here...
Sure, AI will produce a million shitty Google clones, but no one is using them but their makers. Eventually the good ones will start to inch up in users as word gets around, and eventually one might actually make an inroad that Google has to take note of.