NohatCoder

Nah, "selling shovels" is mostly a metaphor, the amount of iron that went into mining equipment was insignificant, beyond some local demand peaks. The majority of the business was consumables.

A key thing to understand about the gold rush is that it was not a major economic event, or at least nowhere near as big as the participants thought it would be, hence the tradegy.

The AI gold rush is different in that there actually is a mountain of "shovels" large enough to flood the global market quite severely.

I guess we can go the long way round. How often does the bakery use 12 million therms of natural gas?

You "suspect", but your suspicion comes with a unit of energy, which simply does not make sense in the context. The rest of us are talking about effect.

First thing missing from this "article" is a number indicating how many people don't want AI datacentres anywhere. You can't claim that someone is a NIMBY if their stance is that the damned thing shouldn't be built anywhere.

I would say that this is a joke, someone made a wobbly UI, isn't that funny. But I get a bit worried when trying to gauge much work went into this that is not required for the joke. It looks like the author actually expect people to pick up this thing and use it, and that is kind of scary.

Yeah, one needs to understand that "unsafe" does not mark which parts of the code are actually unsafe, it simply marks parts where the compiler ignores parts of its rule set. But the implications can crop up anywhere, there is no guarantee that a resulting use-after-free or similar can only happen inside the unsafe blocks.

In short, if you use an unsafe block, then potentially any part of your code is unsafe.

There is no such thing as a complete test suite, there will always be some possible bug that it doesn't catch.

In particular, if you put an LLM in an automated loop of "this test fails, please fix it", there is a pretty good chance that it will simply special case all of the tests, possibly in some contrived way that makes it not at all obvious when you read the code.

If usage of the basics become "bloat" and "a nightmare to maintain", then you haven't learned how to use them.

Have you actually learned to write HTML and CSS without any kit/framework?

I generally don't find that the frameworks help nearly as much as people think they do, and if you haven't learned the basics they easily get in the way of doing so.

"~10k reads+writes per month"

It is the same story over and over again. Your AWS bill is "low" because the actual resources you use are a tiny fraction of what a modern computer can deliver.

This is probably fine for your case, but you could run the same load on any toaster, and without a meaningfully demanding task it is hard to tell them apart.

Don’t forget that people involved in information technology procurement will pay very large sums of the company’s money to not understand anything.

FTFY

No, treating people with hostility and escalating the situation only makes it more likely that someone will snap and attack a cop.

People generally do not shoot at cops, because whether or not they hit the target doing so is pretty much signing their own death sentence. All cops have to do to protect themselves is to not provoke people to fight-or-flight reactions.

The whole WASM thing just went sideways from the start. It feels like the property of being practically unreadable and unwritable for humans was a design criterium, it is certainly the primary "feature" enabled by the stack approach.

Then there is the single array of memory that makes modern memory allocators not really work, resulting in every WASM compiler scrounging up something that mashes the assumptions of the source language into a single array.

I assume that the image would at least fit in L3.

I don't think this is likely to impact development much, they can rent office space elsewhere if needed.

https://nohatcoder.dk/

I don't post often, but I think what is there is quite worthwhile. It is whatever I want to write, but topics are typically maths, game theory and cryptography. There are also a few browser games.

The site itself might also be of note to some people as an example of an extremely light hand crafted website.

Yeah, and I doubt that this is the last price hike. But I don't think this actually has a whole lot to do with ram prices. The thing is, AI cloud providers have generally been running at a massive loss, if you do the napkin maths, the procurement cost of these machines is equivalent to approximately a year of rent at these rates, for normal servers at AWS it is approximately a month.

The nature of these AIs is generally such that you can always throw more computation at the problem. Bigger models is obvious, but as I hinted earlier a lot of the current research goes more towards making various subqueries than making the models even bigger. In any case, for now the predominant factor determining how much compute a given prompt costs is how much compute someone decided to spend. So obviously if you pay for the "good" models there will be a lot more compute behind it than if you prompt a free model.

Lots of things to consider here, but mostly that is not the kind of prompt you would use for coding. Serious vibe coders will ingest an entire codebase into the model, and then use some system that automates iterating.

Basic "ask a question" prompts indeed probably do not cost all that much, but they are also not particularly relevant in any heavy professional use.

So let us compare AI to aviation. Globally aviation accounts for approximately 830 million tons of CO₂ emission per year [1]. If you power your data centre with quality gas power plants you will emit 450g of CO₂ per kWh electricity consumed [2], that is 3.9 million tons per year for a GW data centre. So depending on power mix it will take somewhere around 200 GW of data centres for AI to "catch up" to aviation. I have a hard time finding any numbers on current consumption, but if you believe what the AI folks are saying we will get there soon enough [3].

As for what your individual prompts contribute, it is impossible to get good numbers, and it will obviously vary wildly between types of prompts, choice of model and number of prompts. But I am fairly certain that someone whose job is prompting all day will generally spend several plane trips worth of CO₂.

Now, if this new tool allowed us to do amazing new things, there might be a reasonable argument that it is worth some CO₂. But when you are a programmer and management demands AI use so that you end up doing a worse job, while having worse job satisfaction, and spending extra resources, it is just a Kinder egg of bad.

[1] https://ourworldindata.org/grapher/annual-co-emissions-from-... [2] https://en.wikipedia.org/wiki/Gas-fired_power_plant [3] https://www.datacenterdynamics.com/en/news/anthropic-us-ai-n...

TL;DR:

Defendants Roy and Rianne Schestowitz were the targets of online harassment. They decided that claimant Matthew Garrett was behind it, and initiated their own hate campaign against him, in particular using their websites www.techrights.org and news.tuxmachines.org to do so.

The defendants did a very poor job of going to court, even by the standards of amateurs representing themselves, producing almost no evidence, none of which the judge found to be relevant.

Damages of £70K were awarded.

Wrong, plenty crap make it into the store, that is true for both Android and iOS. And the advertisement in the Android store is designed specifically to try to trick you into installing a different but similar app to the one you wanted.

AWS certainly also does daylight robbery. In the AWS model the normal virtual servers are overpriced, but not super overpriced.

Where they get you is all the ancillary shit, you buy some database/backup/storage/managed service/whatever, and it is priced in dollars per boogaloo, you also have to pay water tax on top, and of course if you use more than the provisioned amount of hafnias the excess ones cost 10x as much.

Most customers have no idea how little compute they are actually buying with those services.

On the contrary, I think you will find that most people simply cannot comprehend the idea that Netanyahu/Israel would aid Hamas, they will think you are a conspiracy nut for stating the plain facts.

While it is good to see some Bayesian statistics in use, I wouldn't in this case put so much emphasis on an exact calculated probability that he did or din not cheat, the prior in this case is simply too wishy-washy for that.

The sound conclusion is that this is not evidence of cheating, but it is not evidence of the contrary either.

This is such a useful feature.

I'm fairly well versed in cryptography. A lot of other people aren't, but they wish they were, so they ask their LLM to make some form of contribution. The result is high level gibberish. When I prod them about the mess, they have to turn to their LLM to deliver a plausibly sounding answer, and that always begins with "You are absolutely right that [thing I mentioned]". So then I don't have to spend any more time wondering if it could be just me who is too obtuse to understand what is going on.

My function approximate coverage of a square pixel, so indeed if you rotate a line around it at a certain distance that line will clip the corners at some angles and be clear of the pixel at other angles.

I do believe that it is already invariant to linear transformations the way you want, i.e. we can evaluate the corners of an arbitrary parallelogram instead of a square and get a similar coverage estimate.

You certainly could imagine doing that, but as long as the initial evaluation is fairly cheap (say a texture lookup), I don't see the extra pass being worth it.

Reminds me that I found an alternative way of sampling an SDF:

First take a sample in each corner of the pixel to be rendered (s1 s2 s3 s4), then compute:

    coverage=0.5 + (s1+s2+s3+s4)/(abs(s1)+abs(s2)+abs(s3)+abs(s4))/2

Relative to the standard method it is expensive to calculate. But for a modern GPU it is still a very light workload to do this once per screen pixel.