miningape

I thought this was going to be the amazing atheist banana clip, was pleasantly surprised to be reminded of this instead

Kamalåso

Simple. Don't produce slop then.

If it offends you so much that people call your work as it is, you should do better work, grow some thicker skin, or stop.

South African food actually has this in abundance, biltong is like beef jerky but better in almost every conceivable way - not to mention all the other variants of it for example dry stix, droevors, etc.

Not to say we don't also have other sweet snacks, but compared to other places I've been nowhere quite has the same level of "savory" snacks

Don't really have anything to add but I do want to say you're not alone - I feel very similarly about AI tooling, the level of satisfaction I get from using them (none), the need for interesting technical challenges, etc. etc.

The raw idea is the only part of this text I found interesting.

Seems you don't like it much when the shoe is on the other foot

I propose "the whack-a-hydra" pattern

> You have a lot of control over what the LLM creates. The way you phrase your requirements, give it guidance over architecture, testing, ux, libraries to use. You can build your own set of skills to outline how you want the LLM to automate your software process

Except for the other 50% of the time where it goes off the rails and does what you explicitly asked it not to do.

> I think that banning is a valid approach

Are we talking about VPNs or phones in general? Because somewhere in this we jumped from VPNs to phones in general, and these things are not equivalent.

A phone ban in general for children, maybe I could agree with you on. But, a VPN ban on those grounds alone is utterly pointless: it's not like there aren't millions of other bits of internet crack children can easily find without a VPN.

Not addressing the actual concerns and instead pointing fingers at a totally different, larger issue reeks of "won't somebody think of the children". All with the convenient downstream effect of revealing which citizens own VPNs.

Yes, as we all know, when evaluating which programming language to use, you should get a line count of the compiler's repo. More lines = more capabilities.

Why would I ever want a language with less capabilities?

Slop Driven Development

It's not about understanding the LLM's theory of mind - the direct quote was

> Understanding the theory of mind that composes the system

i.e. the logic underpinning how the system works

I'm very much in a similar boat to you - I'm also considering a pivot away from SWE if this is what it's going to become. Luckily I'm still young and don't have anyone depending on me (other than myself).

I'm still working on my own small closed source projects, building them the way I want to, like a gameboy emulator - and I've gotten a lot of joy from those.

Or better yet, refactor your app so it doesn't require so much boilerplate - surely if you're doing the same thing over and over again you can just extract it into it's own function / method and abstract over it.

Politics isn't the exclusive hobby of the united states.

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

https://www.deutschland.de/en/topic/politics/the-spd-parties...

And by the way, in the same wikipedia link you posted, there's documentation of Democrats also banning books.

> My anecdotal experience with folks who gravitate towards ideas about unstructured organizations (of any kind) ... also very much expect the output of that organization will just naturally be what they want

Got that exact impression from this article. No doubt the leadership has issues, but I also get the feeling he wouldn't have been happy with this direction even if a group elected it - although, to his credit, it sounds like he probably would've stuck with it longer.

Also not knocking the author, I hope he finds something more fulfilling to work on - and he's probably awesome to work with.

Thanks, I have a pair of nephews this'd be the perfect gift for!

This is really cool

Not sure if they do this, but I'd actually really like if they included instructions for things you can make with just the brick sets - just to help you get started or to come back to later.

fx with a box of 1000 bricks, instructions for 2 or 3 things that can all be made from the bricks in the box (maybe not all at once)

yes this is such a good point, the OG replit could've been the perfect conferencing / classroom tool

Running an IDE in a browser like that is not something I'd ever want to work with long time or experimenting on my "own" computer - maybe it's just me being weird but running the code on the metal I'm holding is much more satisfying.

I'm not sure what features / tools replit had in this regard, but I could easily see it dominating CS education and conferences as the go-to IDE. (then making the real money by monetising the students in the future, i.e. other tools you can sell - even something like replit as a cloud provider), by having features like

  - templates you could share (i.e. one per lesson)
  - live sessions (where the professor could log into many students replit instance and demonstrate)
  - videos built into the editor / streaming / conferencing
  - "homework had-in" features, automated test sharing, etc.

Even the differences between React and Svelte are big enough for this to be noticeable. And Svelte is actually present in the training data. Given the large amount of react training data, svelte performs significantly worse (yes, even when given the full official svelte llms.txt in the context)

The point isn't that you can trick an LLM, but that their capabilities are more strongly tied to training data than context. That's to say, when context and training disagree, training "wins". ("wins" isn't the correct wording, but hopefully you understand the point)

This poses a problem for new frameworks/languages/whatever that do things in a wholly different way since we'll be forced to rely on context that will contradict the training data that's available.

It's well known that even current LLMs do not perform well on logic games when you change the names / language used.

i.e. try asking it to swap the meanings of the words red and green and ask it to describe the colors in a painting and analyse it with color theory - notice how quickly the results degrade, often attributing "green" qualities to "red" since it's now calling it "green".

What this shows us is that training data (where the associations are made) plays a significant role in the level of answer an LLM can give, no matter how good your context is (at overriding the associations / training data). This demonstrates that training data is more important (for "novel" work) than context is.

Because LLMs do not work like that - there's no "understanding" the source and answering questions, it simply "finds" similar results in its training data (matching it with the context) and regurgitates (some part of) it (+ other "noise").

Meaning as technology evolves and does things in novel ways, without explainers annotating it the LLM won't have anything to draw on - reducing the quality of answers. Which brings us full circle, what will companies use as training data without answers in places like SO?

Not a finance guy, so fully prepared to be wrong here. But my interpretation is that an increase in price corresponds to a shorter lifespan. i.e. less time to make money so we need to charge more to get the same return over less time

It could also be a supply/demand issue, generally price increases are caused by either 1. demand increasing, or 2. supply decreasing.

In this case we can interpret a shorter lifespan as decreased supply, but it can also be because the demand for GPU compute has gone up. I think in this case we're seeing a bit of both, but it's hard to tell without more data.

We could also consider the supply / demand elasticity changing, f.x since demand has become more price inelastic it could result in a higher price.

But taking the time to investigate that, get it approved by legal, etc. all takes longer than just slapping a cookie banner component on it.

This is why people complain about the unclear and bureaucratic nature of these laws, it leads to an over complicated investigation and compliance isn't always simple - meaning the safest option is to comply at the highest level and degrade the user experience.

Hey chatgpt, play out a conversation with "Moxley"

I can't tell whether this is dumb or genius.

Yep exactly, at it's simplest all an allocator is doing is keeping track of which memory areas are owned and their addresses. In a sense even the stack pointer is a kind of allocator (you can assign the memory pointed to by the stack pointer and then increment it for the next use)