j_w

How do you "throw away" the bloat when it becomes embedded through the code base? Is GitHub just going to rollback to some commit from 3 years ago?

Commenting just so I can check later what he comes back with.

Simply use a browser plugin (or any other mechanism) that invokes the agent when the "HATCHA" appears.

Number 2 is the thing, right?

Why bother asking? Is a recruiter going to fly out to my high school and see if they can get records? Interview a teacher that still works there?

You also are likely selecting not for high school calculus geniuses, but for people willing to lie. Presumably they ask because there is a more desired answer, so if you want to get a job you should provide the more desirable answer.

+1 to this method. Prior to having a reserved ~6 hour slot of sleep the lack of sleep was getting BAD.

Well they aren't people churning out 40k words :)

I've spent very little time on ao3, but I imagine it's slightly better than RR where seemingly a mark of value is "huge length (word count and chapters)" with weekly updates.

The metro is a much less stressful experience for me with noise cancellation on. Without them the noise in tunnels just makes me anxious. The outside tracks are all fine without them though.

And for walking around - it's the traffic noise that bothers me, not people. Traffic noise can just be so loud along some roads (and at certain times of day) that it makes me not want to walk at all.

I was also confused why commit and push are different steps

I think most people think about file access in the word processor terms, like you mentioned. You "edit," "save," "upload," and "download." Actions like that.

Then they have to use git for the first time and the terms/actions are really foreign. It's likely made worse if they have experience with auto-syncing file software, since that software does the whole remote management process for them.

These criticisms of git always seems so shallow to me.

'add' tells git to start tracking some file(s)

'commit' tells git to save the currently tracked files

'push' says "upload my changes to some other location." Git isn't dropbox magically 'rsync'ing the directory to some server.

'pull' says "download any changes from some other location." Same deal as push.

That should satisfy the majority of git casuals that get frustrated with it. You should learn the tools of your trade, and version control (specifically git) is one of the tools of the software trade. If you work adjacent to software why is it so hard to learn a little about git?

converting all subscription customers to api customers would yield 10x boost in revenue immediately so the demand is there

If OpenAI started charging all subscription users API pricing they would lose the majority of users.

If they increase the price demand will plummet. They have no moat and there are competitors.

Because the same reasoning behind that statements implies that certain races are innately inferior to others. You chose to write "Asians" and "whites" here - why not make the same statement with "whites" and "blacks?"

Saying "Asians" are intellectually superior to "whites" is a thinly veiled way to say "and whites are superior to all other non-Asian/white races."

And the claim that "Asians" are intellectually superior to "whites" isn't even correct "because of race." I'm not aware of any real study that attributed racial identity to measure intelligence. Cultural differences? Socioeconomic differences? Country of origin? Sure. Race? Used as a proxy for the former.

Well the .NET random is bad.

It seems really the problem is twofold: the reference is from 1992 and cites a 1981 publication's reference to an unpublished 1958 generator. Not to say that being old makes the algorithm bad, but it's a bad implementation of an algorithm that already is questionable given more recent research.

I'll go section by section: > //Apparently the range [1..55] is special (Knuth) and so we're wasting the 0'th position.

This is a silly comment. Knuth explicitly states that "24 and 55 in this definition were not chosen at random; they are special values that happen to define a sequence whose least significant bits, {Xn mod 2), will have a period of length 2^55 - 1. Therefore the sequence (Xn) must have a period at least this long."

Then you have the initial seeding of the LCG with with a = 21 and m = 55, which is interesting. Numerical Recipes uses those values, but Knuth whom they got the algorithm from does not suggest them. The closest Knuth suggests is 24 and 55. This suggestion is from 1981, so the viability is questionable (and Knuth clearly states that this is an unpublished algorithm from 1958 - Numerical Recipes itself questions the quality).

Then they use 21 for inextp - this is wrong. Numerical Recipes uses 31, and that is significant per the period length quote above. The use of 21 should measurable lower the period.

Instead if it were a simple LCG using values found in L'Ecuyer's 1999 publication on the topic (https://www.ams.org/journals/mcom/1999-68-225/S0025-5718-99-...) I assume it would have a better distribution.

So the implementation is a questionable algorithm from 1958, and it's done incorrectly. Numerical Recipes opens the chapter on randomness almost immediately with: "Now our first ... lesson in this chapter is: be very, very suspicious of a system-supplied rand()," and then the authors of the .NET random package show exactly why that is.

It definitely takes a lot of work. I've read that it takes a good writer themselves to translate well, since it's such an artistic endeavor.

As somebody who regularly reads translated works, including the occasional machine translation (MTL), they (MTL) suck. You got a hugely biased result, which you recognize.

Translation is hard. If you're familiar with reading translations from specific languages MTL works have a very specific smell to them, it's a bit hard to describe but it's there. A good translation is miles (kilometers, for those outside of the US) above MTL.

That's not to say that perhaps the latest LLMs will have better translation abilities, but that they are generally crap currently. Maybe they are fine for something very short, but absolutely not for longer content.

So if you want them just dumpster dive for them.

And this doesn't even include tariffs...

That's when you bust out the third LLM. Nobody expects the fourth LLM to be the REAL LLM in the chain.

Just live on /active and that stuff doesn't show up.

I don't know why any reply disagrees with this.

Look at basketball in the US. The best players will tell you all they did as a kid was play basketball. You can go to anywhere somewhat populated and the outdoor courts are in use almost all the time school is not in session for pickup play. Outside of structured practice (if they are on a team), many kids are still playing pickup games or shooting casually.

Soccer fields rarely get use outside of structured play. Kids that play soccer in the US just don't play as much, so their skills are (on average) much worse.

There isn’t much of a way I can see to remove data centers from the technological progress we’ve benefited from over the last couple of decades.

That's not really the argument.

The problem with the tweet is that the chart kind of sucks, and it isn't immediately obvious. The category cited is "Computer Software and Accessories" which is under "Information Technology, Commodities"

A more interesting category is "Video and Audio services," specifically the live streaming subcategory. People don't buy software anymore, they pay for subscriptions to services.

Here are links to FRED for both top levels: https://fred.stlouisfed.org/series/CUSR0000SEEE https://fred.stlouisfed.org/series/CUSR0000SERA . Unfortunately the granularity for the Information Technology index isn't available on FRED from what I could find.

So IT price index is down, frankly to a huge degree. But that includes hardware, so it's hard to draw conclusions about software pricing from that specific chart. But Video/Audio services have seen a fairly sizable increase in index in the last decade.

But that's not really very important. We are talking about price indexes, which do tell us roughly how expensive something is over time, but who cares about the price of basketballs unless that's something I plan on buying as a consumer? The BLS charts give a relative importance which we can use a proxy for "how much a price change would affect the consumer." The relative important of the IT category (linked) is 0.745, but the software subcategory is 0.029. Video/audio and live streaming are 0.595 and 0.185, respectively.

Consumers do not purchase software. Companies don't even bother trying to sell software to consumers. The chart linked is tracking a metric that doesn't matter, because it's not important to consumers.

Going back to the relative importance values from 1999 (https://fraser.stlouisfed.org/title/cpi-detailed-report-58/a...) personal computers are at 0.106. No other categories. 2009 (https://www.bls.gov/news.release/archives/cpi_05192010.pdf) is 0.248. 2012 (https://www.bls.gov/news.release/archives/cpi_05152012.pdf) software is included with an importance of 0.048. At that point cable/television is at 1.387 (see Video/audio above).

So software prices don't matter. Consumers aren't spending on software. Service prices do matter, and they are getting more expensive.

This feels like a schizo post. Sorry for the complete lack of formatting.

Oh stop. Both of those are a bit ridiculous. We are talking about the sudden surge of data centers, the past 1-3 years, not the 20 prior to that.

Sourcing a tweet that doesn't have a real source (saying source: BLS without an actual reference to BLS isn't a source) is worthless.

Consumers don't buy software, they buy services. Whether or not that is captured by your source in the "Computer Software" is unknown, since I don't have the source. FRED via BLS has PPI for software publishers going up in the last ~6 years, but that's not exactly analogous to consumer spend.

When you say 2tb for $9 is unfathomable if you travel back in time, yes obviously if you go back to floppy days it's comical. Cloud storage prices have been the same for a decade now (pre LLM boom).

So software may or may not be cheaper for consumers (hard to say, nobody buys software). And in real dollars cloud storage is cheaper because of inflation. Not sure what the significant gain is.

Cheaper software: where is this cheaper software? Most users don't pay for most software. Facebook/Instagram are introducing paid plans now. LLM companies are seemingly starting to increase prices.

Unlimited cloud storage: Where are consumers getting unlimited cloud storage from? I'm not aware of AWS/Google/others reducing prices/providing unlimited plans.

consumers would vastly benefit from cheaper software, nearly unlimited cloud storage, lower property taxes.

What data centers are bringing cheaper software, unlimited cloud storage (for free?), or lower local property taxes?

It's pretty obvious the outrage around datacenters has nothing to do with datacenters and everything to do with knee capping AI progress.

Is it? Data centers are being built with tax incentives given to the operators, no regard for impact on the local utilities, and they bring no local jobs (at least not long term). Seems like lots of negatives to communities if development continues like this.

I hate company acquisitions.

Not for the aquire-ee(?), I'm not going to be a hypocrite and claim I wouldn't take the payout if I were in that position. But that companies can build massive moats by just buying up as many other companies as possible.

I don't even feel like I can make a "good" argument for it either. Massive companies becoming more massive through acquisitions just feels wrong, like the end game won't work out well for the commons.

I assume the point here is that now Cloudflare can try and push more Vite users into their ecosystem? Nudge the development to integrate better with their products? They say they are moving towards Vite, not Vite moving towards Cloudflare, but ultimately <tool> moves in the direction <owner> decides - even if it's "developed openly."

So give real numbers, not "all day battery life."

What does "all day" even mean? It's marketing slop. All work day? All waking hours day? All 24 hour day day?

An Apple release gives you a number of hours and they tell you roughly how they measured it. This just says "trust me it's good." This isn't to praise Apple, it's just the bare minimum.

Rated for all day battery life

So vibes battery life. I'm going to assume it's like most Windows laptops and somehow only lasts 3 hours.

The reality is that most people write horrible code.

The programming knowledge of a university student that just completed their intro programming course is abysmal. The programming knowledge of a university student that just completed the 4 year degree, but didn't spend hundreds to thousands of additional hours working on programming outside of that is abysmal. College classes don't expect you to learn programming to any real extent, they expect you to learn computer science. And the rigor of most schools is even questionable there.

I've been programming for a long time and I'm still not sure if what I write is very good. I know it's better than a lot of what I see, but shiny trash is still trash. I've seen astoundingly bad production issues (bugs are sometimes an understatement) produced by senior engineers. Those people have years of experience and I wouldn't trust them to properly review my code, let alone LLM code.

I do think people should try and learn the basics of any and everything, and I mean everything literally. But if you know the basics of biology are you now able to credibly review ChatGPT's medical advice?

In the section "Everyone should learn some coding":

I would say that the major unlocks are at:

    1-2 weeks: Basic understanding of what the field is about and what general words to use when asking the AI to do something.
    1-2 months: Basic understanding of how and when to ask the AI something.
    4-6 months: Ability to check the output for correctness (using external sources as needed).

4-6 months to check output for correctness? Are we trusting fresh bootcampers in their first week at their first job to do prod code reviews now?

You can learn a LOT in a short period of time, but it would take much more than casual time investment. This is insane advice on the level of telling blue collar workers to just "learn to code."

However, one of the most repulsive ideas is that you can cripple everyone else, because some people have less.

When did I say that I'm in any way pro crippling other students? I'm simply pointing out the socioeconomic reality of school performance.

Comments like yours are vile. Brimming with vitriol.