And with dnf/rpm updates being transactional and easy to undo with `dnf history` or `dnf distro-sync`, I never saw the appeal for a day-to-day OS of fedora atomic & al.: only a worse user experience with a slower system, slower updates, and terrible disk usage in a time of storage scarcity. I keep missing the obvious and telling myself that I'm dumb for it, but OTOH, this box is running fedora 44 with some big COPR (the latest one being for plasma 6.7) admirably well, and the most admirable of all is: this OS was initially installed as fedora 27 and incrementally updated flawlessly for almost a decade now.
I mean, this reminds me of the early imagenet days, when people were first trying to explain the unreasonable efficiency of neural networks at interpreting content in images. They found out that, after enough backprop, some layers of the network became specialized in contours extraction, shape extraction, etc, in ways that can be seen as analogous to earlier CV techniques (Canny/Hough transforms, ...).
Later, Google was having fun feeding whole bunch of youtube content to artificial neural networks, unsupervised, and figured that certain parts of the network would, too, specialize, only to have the activation functions be run backwards and render an abstract image of a cat¹.
None of that is terribly new or surprising for anyone having studied and dealt with neural networks. The only difference today is that the field has completely flip flopped from approaching the subject with scientific rigor and cautious excitement to being a clueless billionaire infinite money printing machine fed on deceiving anthropomorphism and FUD.
> that makes me feel I'm not talking to a self sufficient entity.
I've seen this happening a lot in the recent times: people who are generally not very good at their job tend to offload disproportionately more to LLMs, and it's so damn annoying that their incompetence now comes sugarcoated in lengthy LLM babble for the sake of desperately trying to sound convincing. This is wasting me high single digit hours every week, not to mention the frustration of battling an asymmetrical fight: it takes them seconds to produce something that will take me minutes to read and hours to react upon. This needs to stop.
I spent a bunch of time on a task that we've chatted about for WEEKS.
At the PR review time of this lengthy process, I get a bunch of AI slop saying:
- this looks like it changes X to Y, did you mean to do this? Worth another look?
It's SO frustrating. Just copy pasted BS. Are we really paying someone 6 figures to copy and paste into a prompt all day? This is madness.
I had a deeply maddening experience at one job where somebody was reviewing a Typescript PR of mine. In their first copy-pasted review there was a suggestion that I do something slightly differently; it wasn't something I had a strong opinion on and they were more familiar with the codebase, so I made the suggested change to just keep the peace and get the change out.
In their followup review, again copy-pasted, it made a new recommendation - which was the way I had done it originally. Absolutely no human conviction behind the review, just copy-paste ping-pong feedback.
The way I got around it was by implementing my original change again, and writing a stronger defense of it, with lots of references, and calling out as many downsides to their initial recommendation as possible, in an attempt to prompt-inject and overwhelm whatever model they were feeding my work into, and it worked. It gave me a very grim view of the near future.
I find GBoard to remain superior in multilingual typing, futo just can't seem to be able to switch to other languages consistently, even with the multilingual option enabled,
it's also not as good at "recovering" from typing too many letters (Gboard sometimes adequately completes with likely shorter words)
If you haven't already, you might want to read the twitter thread at the root of this HN discussion. It essentially postulates that all Chinese 3D Printer manufacturers are incentivised to follow in the steps of BambuLabs (being members of the state apparatus and subject to the same rules), while suffocating the competition outside of China (by benefitting from unfair advantages under the umbrella of a nationalistic development plan).
I don't want to sound like a Prusa shill or anything, but if they see Chinese state-sponsored competition as an existential threat (and I see no reason to doubt either claim), then I imagine that, from Prusa's perspective, it's a losing proposition either way ("damned if you do, damned if you don't"): anything they develop in the open directly strengthens their competitor and doesn't elevate the playing field for the benefit of all (unpunished licenses violations) ; anything they keep to themselves turns Prusa further and further away from their ideological stance, more akin to their enemy and less relevant.
For a while (a decade+), I was running CentOS on my servers on the same assumption of long time stability and ensuing peace of mind. Then I figured that over such durations, the ecosystem drift becomes significant and keeping applications up to date and running on top of the OS becomes an increasing challenge (with the more "infrastructure" packages like glibc, python/Apache combos, GCC, ... slowly becoming incompatible with the latest applicative stack).
Then I figured that version upgrades were miserable, not just because I had painted myself in a weird corner with ungodly packages mix-ups, but because the upgrade path was always best-effort. I think I gave up during the 6 to 7 transition, as I realised that all I needed was fedora: with yearly or half-yearly updates I have no need to fight the distro's packages: stuff stays current and in working order, major distro upgrades go smoothly, downtime is minimal. I'm not considering going back to any "server distribution" ever.
You always need to read through the marketing garbage.
Nonethless, in comparision to a lot of other companies, Google has Deepmind and the money to just do all of it without spending money which doesn't exist.
reply