dbalatero

Unfortunately, I don't think AI companies will opt to throttle onboarding new users via a licensing/training system.

Sure, but it's easy to accidentally start up the agent in the wrong directory, like when you open a new terminal. I've done it before when I was distracted (albeit with Claude, not Grok).

Is Fable/Mythos a new generation then?

That's good to know, but I think maybe the bulk of the original point is still concentrated in the "we always need new models to stay relevant" piece.

Of course, this was a hugely poor read of where the costs actually lie in AI. Training - while no doubt capex intensive - is a fixed, up-front cost. You spend hundreds of millions to train a model, then you are "done".

I don't understand this point that people make. If you're consistently needing[0] to train new models and the cost of training relative to the % improvement seems to go higher, isn't this just a constant cost that you continue to bear? The footnote seems to allude to this, but then sort of waves it away anyways. Also are there continuing incremental training costs to keep models relevant? Or do they only have knowledge of events up to the day they were trained?

[0] needing, because you have competitors and people expect more and more.

I type the emdash a lot myself; how do you know the author doesn't as well? The copy is simple and readable so even if it is AI, who cares?

I just ignore people that can't coherently talk about the pros and cons of Vim.

I suspect that while they prefer you to give up all your data, what's even more likely is they are moving fast and breaking things at a rate unseen before, and not enough conversation is happening in design phases where someone can flag that "Hey if you add this new prompt it might break an important user contract you forgot about."

In either case annoying still.

Maybe the difference is more of the professionals in the field now haven't built that same muscle, as there's a broader group of people working in tech. Whereas the folks that could fix things in the 90s mainly gravitated to computers as a profession. Just random musing though I truly don't now.

Do you have it? I've gotten a lot of use out of it, and I find there's a lot of people without it that love to dismiss it in these kinds of threads as useless. It helps me improvise with others, learn things quickly, play in tune on my fretless instrument, to name a few.

I have it, it wanders a bit, but I don't know what you mean by "out of key". If it wanders, everything would relatively wander with it. I've never found it to be a curse in any case.

I have it from early age, and it also shifts now that I'm almost 40. I think the more constantly I use it though, the more it seems to settle back into accuracy.

Regarding OOP itself, I also remember when "favor composition vs inheritance" become a thing. Was this reversed too?

I think this is generally still the advice, when working in OOP contexts.

Yes, and/but, many enter life once the horse riding is the dominant culture. It never even crosses their mind that smelling the roses is a thing they could or should do.

Hey I mean AI is supposed to make us 10x more productive, so the price of things should also go down 10x right?

...right?

Do you have access to their financials?

Is this in things like the clothing industry, where there exists a conversation around fast/cheap/outsourced fashion and has consumer pushback built in? If so, it makes sense they would get ahead of that. I'm not sure all industries bother to make that point/consumers really care.

I could be wrong, but it feels like one issue is that AI seems to cater more as a signal to venture capital and the internals of the tech industry in a lot of these products, while consumers just want to know "what is this product going to actually do for me," and care less about whether it is implemented with the buzzword du jour.

I think typically it's karma farming? To what end I'm not sure but it's probably not great.

When I try to share this page on iMessage it unfurls the link as "LLM / AI Standard Test Page" so I guess I won't share it!

How do you see it getting sorted out over time?

I thought they were all in on agentic coding? They are probably just building at a surface level with only an eye towards shipping, without considering the impact of all the changes. I've seen less and less coordination between engineers as well under that model. If that's the case (Claude Code is this way). it is sort of what you get, no matter the rhetoric about "make sure to review all your changes!" It's always trade offs.

Hopefully the migration function doesn't silently move files. If you have dotfile symlinks, those would need updating. Also it would be surprising if config files were in a new place one day.

This companion piece was also interesting to read: https://nlkw.de/en/blog/vocabulary-of-getting-fried/

But I'll be long dead before that plays out.

Given how that's played out in the movies, I'd be happy about that.

I've personally seen way more of a bias to shipping something because "now it's free" without actually discussing whether it's worthwhile. "Just do it" and we'll measure it later. Often the later doesn't happen though. I think this article is a good reminder that applying taste and asking questions is still valuable, even if the code is considered to be cheaper.

Sure, but it's only a shakedown if it's an unoccupied second home, which is hard to have sympathy for. It can easily be an occupied second home (family, renters) or a first home for those in the upper middle class paying for $1mm+ apartments in NYC. I'm not really worried about Jeff Bezos or some Hollywood actor's crash pad when they have business in nyc.

I mean it was always easy enough to say "Hey not quite sure but I did find this post on SO, in case it helps"

I was mostly making a joke.

Did you mean AGI? I'd think whatever else they are doing, LLMs are reducing RSI...