ℍ𝕂-𝟞𝟝

I guess it makes coreutils vulnerable to EEE. Corps with significant stakes in Linux would be able to extend the API with proprietary functions, build ecosystems on top of that, and lock out independent development.

Imagine being unable to distribute a Linux based OS that is compatible with most of the world without proprietary age verification built in for example.

Are not all ORMs like that? I only used ActiveRecord before fucking off from backend 10 years ago

I actually did an experiment on doing just that. For context, I’m an experienced software engineer, whose company buys him a tom of Claude usage so I had time to test out what it can actually do and I feel like I’m capable of judging where it’s good and where it falls short at.

How Claude Code works is that there are actually multiple models involved, one for doign the coding, one “reasoning” model to keep the chain of thought and the context going, and a bunch of small specialized ones for odd jobs around the thing.

The thing that doesn’t work yet is that the big reasoning model has to still be big, otherwise it will hallucinate frequently enough to break the workflow. If you could get one of the big models to run locally, you’d be there. However, with recent advances in quantization and MoE models, it’s actually getting nearer fast enough that I would expect it to be generally available in a year or two.

Today the best I could do was a tool that could take 150 gigs of RAM, 24 gigs of VRAM and AMD’s top of the line card to take 30 minutes what takes Claude Code 1-2. But surprisingly, the output of the model was not bad at all.

Standard TLS, I think, but what else would you need?

In Nginx you can do rewrites so services think they are at the root.

Oh, there are those as well, I’m not dunking on juniors.

It’s just that my problems always tend to be caused by mismanagement of people.

Like just today I had to clean up after a “let’s do a quick and dirty experiment, oh it works so now it’s production, make 200 more features in a month built on top of the quick and dirty let’s just try it code, what do you mean we lost millions because of a regression nobody even noticed” situation.

Yeah that shit is more common than people think.

A big part of the business of cloud providers is that most orgs have no idea how to do shit. Their enterprise consultants are also wildly variable in competence.

There was also a large amount of useless bullshit that I needed to cut down since being hired at my current spot, but the amount of containers is actually warranted. We do have that traffic, which is both happy and sad, since while business is booming, I have to deal with this.

I know using work as an example is cheating, but around 1400-1500 to 5000-6000 depending on load throughout the day.

At home it’s 12.

The people who would be okay with this already don’t own computers, they go with a phone.

rmdir ./~

Couldn’t we buy from Taiwan? Or better yet, build our own devices with Taiwanese processors. It’s much harder to spy if the supplied components are simpler and the device is opaque to them but transparent to us.

If you look at what AI does, however, it’s mostly classification.

Not necessarily, a huge use case is regulation and control in the engineering, not the political sense. Like driverless cars, independently flying drones and such. And yeah, they need classification subsystems under the hood to work, but their ultimate outputs are complex control signals, not simple classes.

And don’t get me wrong, I also like ML and AI as a field, I just don’t like how OpenAI fucked the field with text generators that they got Silicon Valley to worship like gods. I even like LLMs, just not the grotesquely outsized cult around them.

AI is an adequate solution to a problem that has no other similarly adequate solutions (classification of complex information).

Sentiment analysis machines and such have been around before LLMs and eat much less electricity.

LLMs taken over the “AI” label so much that any success from a machine learning context is attributed to it, while it actually defunds and kills research out of ML all into LLMs.

Excel used to have, and I think it still has, localised function names.

Makes it a nightmare to look up stuff on the Internet.

With a JS-based query language, yeah

I’ve a slight manageable case of ADHD and I tend to obsessively hyperfocus on tasks. It’s a good relationship because I get a lot of shit done well, and enjoy my work.

If you start forcing me to plan out my day every day, down to 15 minute increments, my productivity drops by around 60%, because I stop concentrating on getting shit done, and start working to rule. Not because I’m vindictive, but because that’s what you asked me to do.

So they got their feelings satisfied with only a major annoyance to everyone and about a month of work wasted among everyone.

Definitely, and I’m saying that while my jobs were mostly on NoSQL and I love doing it.

My point is more that 90% of use cases don’t need that, and for those that do, you can’t just slap eg. Cassandra at it and pretend it’s a relational database.

MongoDB is huge though for all the wrong reasons, businesses think that just because it’s JS, they can just have frontend devs - sorry, they are “fullstack” now - doing DBA work.

I worked as one of two NoSQL DBAs for a Fortune 50 finance company, and there is a ton of CV-driven development going on giving NoSQL a bad name. Most use cases don’t need NoSQL. And for those which do, NoSQL is almost always harder to implement than simple SQL based RDBMSs.