rpdillon

27% of crops grown in the US are used to feed humans.

I read a book called "Nothing to Envy: Ordinary Lives in North Korea" by Barbara Demick, about North Koreans in the 1990s that escaped into China and told their story. One of the interesting parts of it is about the planned economy, and all the control the government exerts over the economy in NK.

There was a massive famine I didn't know about in the mid-to-late 1990s in NK, and millions died. During this period, an underground capitalism took over, where people were producing food in various ways, often grown in their own gardens, and then buying and selling from each other, with barter rates / prices determined by availability.

While I agree with you that countries can make rules that fix prices (or do whatever they want), there is an underlying set of mechanisms to supply and demand that are not derived from law, but logic. If I have 5 cupcakes, and there are 100 people that want them, unless someone points a gun at me and says "Sell them each for 5 cents" (while the 100 people all fight over who gets to pay), the natural incentive structure would essentially be an auction to figure out who is willing to pay the most, and then sell it to them. That's not some legal theory or set of laws, but rather the human condition.

Of course, it doesn't play out this way in reality. In reality, it all comes down to power, and those with power take whatever they want, but my point here is that supply and demand exist outside of a legal structure, and if there's only one electrician, and there's enough work for 10 electricians, that one electrician can server the richest 1/10 of those in need (and charge accordingly to select that 1/10), and it is rational to do so independent of any laws made. So much so, in fact, that laws are advisory, but the fundamental incentives are inherent.

This is why I like incentives more than rules: they're less brittle.

> Qwen 3.6 27b locally a few months ago

I've been doing local inference for a couple of years on the side, and I'm astonished at the number of variables you need to have control over to get a reliable result. Inference engine, model parameters (top_k, temp, MTP-enabled/not), quant level, and harness all have a big impact on the results.

DS4-0731 at 2bit on llama.cpp (ROCm) and 250k context with omp.sh has been consistently reliable for me, just a bit slow (10 t/s) compared to what I'd prefer. Trying out DwarfStar today (benching it right now) to see if I can get better speed, but otherwise I've found it to be great on my side projects that are smaller (up to 10ksloc).

There could also be a domain issue - I tend to do lots of web programming and sysadmin work in these projects; if your work is more esoteric, it might not be nearly as good. I haven't tested much outside of my narrow domain.

Quants vary by model. DS4 is very credible at a 2-bit quant. Not sure about Qwen 3.8 Flash Next; I run it at a 4-bit quant and it's too slow, so I'm trying out DwarfStar today to see if that improves things.

> how are they going to prove whether someone understands the code or not?

By discussing the code.

> maintainers can simply sabotage them and accuse them of using an LLM to explain the code

Bad faith enforcement is possible no matter the rules. If you think it's bad faith, a different policy won't save you.

> question the competence of the llama.cpp dev team

Following them for years, they seem extremely well put-together, and have excellent judgment. They are using the same policy as Linux and Debian (in my words, the speed of light is human understanding and judgment). Whether it is reasonable is a different question from enforcement, which typically comes down to "this seems fishy, explain your reasoning".

As for code review, the rule of thumb I've used for decades is: it takes about as long to review and understand as it does to write. Your 100x metric is completely outside of anything I've seen in any hobby or professional project, ever.

I'd like to see specific files you scanned and specific vulnerabilities cited.

Yeah, they're adding all this functionality to the hardware, as though that's the reason that I'm not using Kindle. Now that they've locked down the ecosystem, I only pay for DRM free books.

I think both Gates and Obama said that there's a non-zero possibility of it, but that it's not what they're worried about. And I agree with them. I think the fears are vastly overblown because both OpenAI/Anthropic and the media benefit from the explosive narrative.

This is the way. I do the same, but it's because I don't care about monetizing anything, I just like making stuff.

Yep. Mafia Wars was what led to to quit FB in 2009. Writing was on the wall.

But security is context-dependent. How often does HN log us out?

That solves "random" but not "I don't wanna type my password every week".

Interesting:

> Two changes are urgently needed. First: AI companies need to rely more on the safety expertise that already exists in other fields. And second, before we create systems significantly more capable than the ones we have today, we need new science to ensure that more capable models (and their successors) will make safe choices when we aren’t looking.

He mentions farther down about learning from aerospace engineers and nuclear engineers about safety. Those industries are heavily regulated, so perhaps regulation above a certain capability level is needed. Defining what that level is might be tough, though.

The second point is harder: in the field of AI, practice has extended far beyond theory, so his call for new science is going to be fundamentally tough, because we can't effectively coordinate a global slowdown in AI development so we can let theory catch up. This means, like so many other industries, the safety lessons will be written in blood.

Agree, and both are on the front page. Also, Show HN is usually for your own project, but this appears to be a blog post about someone else's project (albeit "a good friend").

This doesn't seem to be true. There's a clear legal path via the first-sale doctrine to train models on copyrighted works. It's been years now, and publishers still don't seem to be offering anything for training (e.g. bulk licenses solely for training use), but adversarial interoperability via cutting up books and scanning them remains perfectly legal.

There's also the ability to distill other models, which is also not illegal (though I'm sure they like to come after whomever for TOS violations, but thats a civil matter).

And, of course, the obligatory copying-isn't-theft observation. A recent supreme court judgment put it well.

> Since the statutorily defined property rights of a copyright holder have a character distinct from the possessory interest of the owner of simple “goods, wares, [or] merchandise,” interference with copyright does not easily equate with theft, conversion, or fraud. The infringer of a copyright does not assume physical control over the copyright, nor wholly deprive its owner of its use. Infringement implicates a more complex set of property interests than does run-of-the-mill theft, conversion, or fraud.

Folks are pretty smart here, I think we can handle these nuances, even if we don't agree about whether they are good.

Edit: reading through the full text of their post, it looks like they are using common crawl, which is likely just as much of a copyright infringement as Anna's Archive -- it's not like published works have a unique claim to copyright. I think this strengthens your point, though: I was expecting to see scans as training data, but it doesn't appear to be the case.

Software from before the current era, where we thought giving the users choice was a good thing. Now everyone gets a bland default, and people no longer invest in, or particularly care about, their tools, since some PM that has no idea about your use case will just change it in the next update anyway.

I don't mean to be cynical, but the trend is very real, in my experience.

Ha, been avoiding those like the plague. Thankfully still passkey-free, but I can see that the day is coming where someone's gonna force it. Hopefully Keepass has my back here.

Yep, the decade-long timeline about matches what I recall. Agree on the culling that's coming in the AI space as well. Temporal's comment from a couple of days ago about inside vs. outside AI has been thought-provoking as I think about who is going to go bust...

Been a huge fan of QR codes since the beginning. URLs are awesome, and QR codes are a way of putting links in the real world. Obviously awesome! There was a dark period in the early 2010s where everyone I said this to sent me a link to an empty tumblr "people scanning at codes", but I'm glad they've gotten more traction. Took a trip to Santiago this past week, and they are ubiquitous: from room service and connecting to WiFi to menus in every restaurant. Super cool.

Yes. UI design is a real skill, as evidenced by the majority of UI designs being frustrating and confusing for anything but the most common use cases. AI interfaces to all these systems isn't as good as a great UI, but it's a lot easier to make and it's a lot better than a bad UI. I suspect AI interfaces are going to become the "worse is better" of the 2020s.

The point is about trajectory. Apple continually removes rights from developers, and then users, but grants all those rights to themselves. This has been going on for a while. I left the Apple ecosystem in 2011 because I was concerned about the direction, and in the intervening 15 years, I have not regretted it. The amount of random breakage I get on Mac is annoying enough that I'm happy to get back on Linux.

Didn't the overbuild of fiber directly lead to Google buying up dark fiber, and then lighting it up later on to power YouTube? I ran across this when I was trying to figure out why YouTube has no real competitors, and the explanations I ran across were basically “Google has an immense economic advantage because of their control over all this extra bandwidth capacity". I have not further validated that in the couple of years since I read it, but it does make me wonder what excess AI capacity could lead to.

This is very true. I'm more careful with what I ask my local model, since it's about 1/3 as fast as the cloud models I'm used to.

Totally true. But then it introduces a much greater degree of control, and it creates the aforementioned local concentration of energy usage, noise, etc. So it's a tradeoff, more than a strict win.

Similarly with my Strix Halo running DSV4. Got a meter for it a couple of weeks back, and, during inference, it's pulling about 160W, and I run it about 15 hours a day on a busy day, or about 2.4kW. I'd say an average day is more like 2 hours for me. It's a small impact on my monthly bill. Owning a hot tub is orders of magnitude worse, in my experience.

How about phone battery manufacturing? Shall we do an accounting of abuses there? Phones warp minds, and have extensive slavery overseas in the supply chain. Societies love to claim the moral high ground by externalizing bad behavior, and importing the result. The US does this extensively with climate and labor laws, and Europe does it with pesticides. The difference with porn is its pretty girls, and its all on camera. I see some parallels with Missing White Woman Syndrome.

https://en.wikipedia.org/wiki/Missing_white_woman_syndrome

The efficacy there is going to hinge on step 3:

> Select Transfer to see if your order is eligible for transfer.

Anyone have any idea how often events are allowed to transfer? I assume the artist/venue drive this decision.

I don't use the app, I just use the website. Then visit:

https://www.amazon.com/wholefoodsapp/prime-code

The web was smart. URLs were smart. We've lost a lot moving to crappy user interfaces people have to dig through. I'm have hopes this will get better soon.

Edit: Isn't there trivial fix with scanning a QR in the store that either opens the site or deep-links into the app? This is like 15 minutes of store manager time + one Claude prompt away for the mobile app, I'd think.

I was going to say, I scan screenshots of passes/codes all the time, including airplanes and busses. I haven't tried a photo, but I feel like there's a good shot it would work just fine. I wonder if a gallery app could be a poor man's pass app.

This is closer to the reality I'm seeing. Software engineers are well positioned to understand and deploy LLMs to solve problems, even in adjacent domains. One of the engineers I work with gave a short presentation yesterday. It was quite simple. He'd just taken a series of screenshots from our product and put it into a Claude desktop session with some pretty significant prompt explaining how the screenshots related to a particular user flow and what the desired changes we were thinking about making were.

Took about 45 minutes and produced a working mock that we could then play with and see if the UX felt right. It was surprisingly high fidelity, and while the code was entirely throw away, it helped a room full of people make decisions about how we wanted the tool to feel within the first couple of hours of thinking about the project. Designers bring a wealth of UX thinking to that discussion, but the process of creating the mock was done entirely by engineering.