vincent-uden

I feel similarly. At some point in the past, many static type systems were not expressive enough to handle some of the useful strategies available to dynamic type systems.

But nowadays, pretty much all static type systems employ some (or multiple) convenient methods of polymorphism

As you mention in your post its more about keeping up with the moving bleeding edge rather than keeping what you have. The most important point/habit is to be invested in lifelong learning.

The moment you stop learning, only doing rote work, is when you stagnate. Thankfully I don't know almost anyone in software who can do even a days worth of work without something unexpected happening. As we all know, execution of a plan is never as simple as the plan. Each unexpected moment is an opportunity to learn, make sure to squeeze those moments for learning and knowledge as much as you (and your employer) will allow.

If your job doesn't present many opportunities for learning then the answer would be to look for another job or hobby-projects if you have the time and energy available.

Always has been. Seriously, has Lunduke ever said anything in good faith without coming off as a smug asshole?

Charging extra for it at least lets me ignore it :)

Yes, the experience here on HN, like any open software forum has been watered down for a long while now. Before you could respect even quite simple projects since good time was at least spent implementing them. Nowadays? Not so much.

The "S" in "IoT" stands for security

One year after publishing my sort-of-release announcment of my PDF reader, I'm back to discuss the maintenance of this Free software.

The "critically acclaimed" original article is of course still available at https://www.uden.dev/blog/pdf-reader

Also, the release notes for 0.12: https://github.com/vincent-uden/miro/releases/tag/v0.12.0

Not at all. There are way more valid lifetimes than the lifetime system of Rust can express. There are countless scenarios where you could accidentally use the unsound behaviour while not producing a segmentation fault because the lifetimes happens to be arranged such that no invalid memory access actually happened.

A program isn't invalid just because it violates the ownership and borrowing rules of Rust. There are tons of programs in the world with shared, mutable access, devoid of Mutexes, Arcs and Locks that nonetheless are completely valid.

What the hell. This is not the esoteric, crazy, impractical constellation I expected to find with a title like this. The minimal example is so simple and is definitely something you'd easily come across in normal code.

> We are in fact seeing software improvement, but the "we" is very limited. Most people don't adopt new software very often, so it's not surprising that they aren't finding new great software now.

This is just plain wrong. I don't think the people asking for new software are in any way averse to adopting new software. I for one am actively waiting for new software to be released so I can adopt it, the Rad debugger for example. The first Linux beta-builds have been shipped and I can't wait for it to be in a more usable state.

I try new software all the time, not to mention I usually try the same software multiple times before reaching a final verdict. In the text-editor space I've been around Vim, Neovim, Doom Emacs, Spacemacs, Zed, Vs Code, Helix, Kakuone, all of them several times over before settling on Neovim.

I would love for more actual CAD-alternatives, more Photoshop alternatives, more video editors (shoutout to Blick) that all cover some niche. They don't even need to be full replacements!

But no. AI has not helped on this front in the slightest. If anything the bleeding edge (if you wish to call vibe-coded, half-assed clones that) is in its most broken state ever.

I see. In that case you could see social media as a one-way street. You post your articles and you don't have to go back and read or interact with comments on the respective site. You could even use some crossposting solution to post to all the medias at once (I think these exist still).

I suspect growth will be slower if you don't interact at all on these platforms, but by the sounds of it that would be an acceptable trade-off for you.

On my biggest post/article I did not have the time or effort to respond to even 10% of the comments here, and that is/was totally fine. The article and its comment section on HN had a life of their own which I was not required to partake in.

Linking to your blog on your profile here is a starter :)

But seriously, it is kind of difficult and it depends on your definition of social media. Is Hacker News a social media to you for example?

I've had some great success in the past just posting my articles here. As you can see from my post history I'm not exactly "active", posting 6 times in the last year.

There is always the traditional routes of paying for targeted advertising, email campaigns and so on. However this depends on the type of "growth" you want. Does your blog have a comment section? If not you kind of need to post it somewhere for you to get any feedback whatsoever. If you are writing solely for the purpose of advancing your own career I struggle to see why the articles would be interesting to read for others. So I at least suggest having some other motivation than that.

I realise that my answer contains a lot of "ifs" but it is really hard to answer such a vague question without any particulars.

We all know that the gain in "efficiency" produced by this new technology won't be distributed to the masses, as with any technological increase in a capitalist system.

While this particular boulder might not have to be pushed by the same, large amount of people, the same amounts of mouths still need feeding. The Sisyphean task of surviving is still far from (if ever) self-pushing. Absurdism is still as relevant as ever, if not more relevant than ever before.

I'd even go as far as to say that the revolt against the absurd is one of the philosophical ideas that can help one go through this apparent loss of need for boulder-pushers. To revolt against the absurd is to create while perfectly aware that the creation itself is meaningless (which I feel like any pure mathematician should be aware of and have already accepted due to the nature of their work). Of course, this is separate from the question of funding, salary and the realities of life. But I feel like mathematicians and other scientists aren't just freaking out over their financial situation.

> Idiot says idiotic statement

That is partially what I do! I don't favourite stuff at all really on HN/X/whatever. I use https://obsidian.md/clipper to download them straight into my vault.

Then there is https://pi.dev/packages/pi-obsidian which I use to give an agent access to my vault for searching/organising tasks

I use obsidian with obsidian sync to keep track of most things. Among these notes I have tiny notes which are basically just a title and a url, possibly linked to some concept or other note.

Easily searchable and navigable too.

If you want help you could start off by listing what you know about the two. Do some of your own research and ask concrete questions about the two frameworks.

There is such a vast difference between most decisions of these two frameworks that it is really difficult to even know where to begin with an answer.

So, what do you know about them?

Interesting! As I said I've only heard complaints from Americans since I have no insight into how the tax filing works. If just general Claude is capable of doing it, then that is just as satisfying of an answer to why there isn't specialized AI tax software.

For starters, the average person doesn't use LLMs in any capacity beyond "ChatGPT" or whatever chatbot interface they happen to have latched on to.

I don't know the financial penalties of a faulty Tax filing in America since this is just done automatically for individuals in Sweden by our IRS-equivalent. But I assume this is something you don't want to mess up. If so, apply an general stochasticity argument for why AI isn't end-to-end responsible for much at all where details and correctness really matters.

Yes they are. Just because the models have been overfit on the entirety of the internet doesn't mean they aren't in principle re-generating parts of their input data for the overwhelming majority of their output. Sure, stochasticity pushes the output into weird states sometimes, these are the states where the output is plain useless nonsense. Hallucinations, if you will.

Our brains are entirely too small to recognize the patterns encoded from the gigantic training data set. But they are there nonetheless.

This is easily seen if you try get *any* neural network to generalize to *any* non-trivial task (in this case, XOR for example I consider to be trivial). There are no non-flickering, stable vision models that can recognize a human/dog/cat/whatever in all the lighting settings we can without false positives. The models are only getting better because they are getting bigger and are provided with more data. This is *not* generalization.

You actually have no idea if an LLM works "a lot like the human brain". No one does. It is still not clear if neural networks are capable of generalization to data outside of the training set in any capacity. Not just LLMs.

If they could, why would we need these monstrous models, even for simple problems?

For training purposes? I'd say no.

The legal or ethical status of generated output is still up for discussion. I myself do not think neural networks can generalize and thus I would have to say that it still would be copyright infringement since its just regurgitation of input data. Just because I bought a book doesn't mean I get to re-publish it.

I for one am still baffled at just how OpenAI/Anthropic and the likes have gotten away with the biggest copyright infringements (or perhaps IP-theft, I'm not sure what the correct term is) in history (the data collection for training) without any legal ramifications.

Even if it turns out to be perfectly legal I think a "retreat" to this position of LLM generation being Plagiarism (and therefore being uninteresting and morally bad) is quite a reasonable action.

It also separates LLMs for searching through documents quite nicely from content generation which since I would be hard pressed to consider a "fuzzy search" as a form of plagiarism.

Yes, that should be evident from the first paragraph of each article.

This isn't at all the title of the article? The actual title is "AI Solves a Major Unsolved Math Problem. Not Everyone Is Happy", I was left really confused.

I don't quite understand what the Pdfium dependency is doing for this project if it can render PDFs on its own. Is says its for native rendering, but what is the non-native alternative this project, written in Rust would use otherwise?

Nope. I find that almost anything I do with an LLM leads to me having a lesser understanding than I would have otherwise.

I use it for search through big documents as a sort of fuzzy finding or sometimes a summarizer, so in that sense I use it to sift through research papers. For understanding or learning, nothing beats struggling through the work yourself. That is the only way for me to make information stick somewhat permanently.

I' ve heard so. But I'd still consider my opinion anecdotal because I haven't done any research into the subject (or measured it myself) really. All I've done is tried both harnesses.

I feel like the models work better in the more minimalist harness. It's totally anecdotal but I feel like the models get lost whenever I run them in OpenCode. Sort of like the "10x"-engineer who overengineers everything.

I mean, kinda same. I've asked it to write skills but that is about it. And skills dont require this. But I do use other peoples extensions.