Notices by ⛈️ Information ⛈️ (elucidating@mastodon.social), page 3

Also, a lot of apps have regular updates. Folks expect apps to have some sort of active life, and even judge products or projects as "dead" if they aren't altered. To help facilitate this, many use self-updating systems.

Nix can't really tolerate this. It's hardly unique here, btw.

Recently an old acquaintance came to me expecting that I'd be sympathetic to their extremely awful AI grift, because they saw me saying "the technology has potential."

I confess, I hesitated before telling them it was not good (I used stronger words). I feel bad about all this, but I really mean the best for folks when I tell them:

You cannot build a career in tech superficially chasing fads and the latest hype cycle. There is no substance there. You will end up a hollow grifter, as well.

And I think this is where a lot of people in tech who marched through the world of startups (especially those like myself with any degree of success) need to recognize: the 2005-2020 era was basically a lot of people participating in ponzi schemes, in which the bagholders were novel: tech workers accepting private equity with dubious terms in exchange for the privilege of executing the "mission," namely their incredibly time consuming labor to make these products on the off chance they blow up.

This is *especially* obvious when you meet a lot of "AI engineers" and realize that if their underlying API doesn't support it, they can't do it and you realize, "Oh... you're a python programmer who's read the docs. You're... you're not actually an AI specialist. You know this Python API."

There are obviously enormously smart, respectable people in the field. But they are not needed or even wanted to make a customer support chatbot or copyright ripping sextbot or whatever. That's for the marks

Lots of people go into tech, they get a positive signal from being the technological equivalent of an agreeable party person, and never gaining any technical depth, leadership skill, or experience actually shipping a project.

And then they're in their late 30s or whatever, realizing that they're lying to people about their finances while waving around an expired founders card. They're living out of an illegal pod complex in SF's Mission district and they have no prospects, no real career.

The way a lot of folks talk about "accountability" in AI is kinda scary sometimes. Like the primary goal is to have someone to punish for a mistake rather than going, "Oh this is a systemic problem."

I would simply not build shitty systems, but the problem is this same strategy often indicts purely human systems and this is weirdly unpopular in tech discourse.

I don't understand what kind of mystical power Nix has to make good, smart people passionate about a broken, janky, compromise-riddled thing the way it does.

I'm increasingly annoyed at these big long passion threads about Nix that diverge so wildly from my reality.

I've had nixos permanently break hardware (fingerprint sensors on my framework are permanently unusable), Nix flakes has been trying and failing to fix problems for years, and still no one knows how to use it. It's got a massive dickbag for a creator and a broken community but folks talk about how helpful and lovely everyone is.

@mekkaokereke Completely agree. I wanted to add some rays of positivity for those feeling bleak, I hope you don't mind:

It's worth noting that while the US government under-invests, it's not entirely bleak for the US green tech scene.

Specifically, both Iron-Air and Sodium-Ion batteries are US inventions and even manufactured primarily here today. US companies are also much closer to cost-effective LAES than most EU or CN entities.

That all changes if we don't push harder, but it's a start.

@mattly @danilo I think this is a good point and it's why I've been trying to tag more, but also it's understandable that folks in a panic might also have poor discipline about CWs.

One of the most fascinating things about modern society is that we can, correctly I think, no long assume elected government actors are dramatically better informed than the populace.

In fact, it's increasingly apparent that this was always a function of the administrative state, and only appeared as omniscience via faith, confirmation bias and a belief a government apparatus can work.

Social media has slipped the lead.

*rubs temples as I handle this complaint at work again*

Look, if you're comfortable thinking about the average of a value over time, the *same assumptions* justify a normal, beta or poisson distribution for that value over time. And with some modest degree of computational analysis, it's always more informative to model the distribution of the value rather than just the mean.

Clojure's approach was quite good. But it fell out of favor for... you know... reasons.

Even Hickey seemed to walk back from it.

It's extremely depressing that the industry turned away from software transactional memory.

The argument deployed to me was always more consistent p99's. Of course, in reality the lockfull implementations are way less consistent because general lockful programming is extremely hard. But the unit tests tell a different story, and the industry at the time was extremely about "test first."

And most organizations are like "well our SLO is six nines under this target, so we can't use a tool like STM on principle" but in reality, very few organizations ever actually live up to any more than 3 nines on any sort of latency SLO, even if you scrub all external network jitter.

So Trump just told a ton of his supporters, verbatim, "I don't care about you. I just want your vote."

And that's in response to concerns people were leaving his LV rally during a terrible heat wave that was making it difficult for the crowd to stay?

I like to pride myself on believing I have a working theory of mind for the conservative brain, but every time I think I got it a new kind of abuse is endured and I have to figure out how adults take that kid of abuse.

I wish that the story around masking was simpler, but it isn't.

This isn't to say you shouldn't wear a mask. Please do! But you should be clear eyed about what your masking protocol can do. Particularly around COVID, which has gone from "very infectious" to "some variants were more infectious than measles", we have to think about environmental tools as well.

And also just risk assessment. Don't get on a train car with a person who's obviously sick, for a trivial example.

I see folks underestimating the challenge of a serious masking protocol, probably because they're unaware but possibly because that reality is very depressing to contemplate.

For example, people with facial hair may need to shave more than once a day to maintain a masking protocol. If you train for disaster assistance they discuss that, but approximately 0 hours impassioned online disucssions of masking link ANY resources on how to maintain/check a good protocol. Check my reply for one.

The more I think about it, the more I laugh. The reason AI maximalists are so eager to imagine a world where an attention model can compress most of human knowledge and reasoning is so simple and selfish:

They don't want complex back ends.

Think about it, if these systems really worked as advertised they'd be a grannd unifying theory of software. A total revolution. Only AI skills and systems would matter.

And that's why so many of them struggle so hard to ignore the growing consensus.