If the code doesn’t compile, or is badly mangled, or uses the wrong APIs / imports or forgets something really important then it’s broken. I can use AI to inform my opinion and sometimes makes use of what it outputs but critically I know how to program and I know how to spot good and bad code.

I can’t speak for how you use it, but if you don’t have any real programmers and you’re iterating until something works then you could be producing junk and not know it. Maybe it doesn’t matter in your case if its a bunch for throwaway scripts and helpers but if you have actual code in production where money, lives, reputation, safety or security are at risk then it absolutely does.

I have never seen an AI generated code which is correct. Not once. I’ve certainly seen it broadly correct and used it for the gist of something. But normally it fucks something up - imports, dependencies, logic, API calls, or a combination of all them.

I sure as hell wouldn’t trust to use it without reviewing it thoroughly. And anyone stupid enough to use it blindly through “vibe” programming deserves everything they get. And most likely that will be a massive bill and code which is horribly broken in some serious and subtle way.

I’m sure declining viewers has nothing to do with the various controversies such as auctioning off prototypes, rushed reviews with misleading or false conclusions, mistreating staff etc. Channels like Gamers Nexus really laid into him.

What isn’t made clear is if this had anything to do with him being a LibreOffice developer, or just the usual Kafkaesque bullshit that happens when someone’s account gets flagged and they cannot get a real human being to appeal or help reverse the problem.

Did you read the text? This guy was providing a package because the default one was broken and he’s fed up of dealing with complaints. And the solution to that is just flatpak the thing and tell users to use that regardless of dist.

The answer for this guy and other people stretched by supporting Linux is to say it’s flatpak or nothing. Stop trying to build for each dist because it’s not sustainable. If someone on a dist wants to maintain a package then let them take the heat if it is broken.

It would have been smarter for the UK to mandate that every ISP must provide a family filter for free as part of their service. Something that is optional and can be turned on or off by the account holder but allows parents to set filters (and curfews) if they want. They could even require that ISPs require new signups to affirm if they want it on or off by default so people with families are more likely to start with it enabled.

While I wouldn’t want flakpak going deep into the OS I think the advantage of using them on the desktop is obvious. Developers can release to multiple dists from a single build and end users get updates and versions immediately rather than waiting for the dist to update its packages.

The tradeoff is disk consumption but it’s not really that big of a deal. Flatpaks are layered so apps can share dependencies. e.g. if the app is GNOME it can share the GNOME runtime with other apps and doesn’t need to ship with its own.

I’ve set my P1S to print through LAN and it works fine. I don’t want or need to use an app to control the printer so I’m not concerned by that loss of functionality. IMO printing via LAN makes more sense anyway for most users. Having to upload some multiple megabyte file to the cloud just to download it back down to a machine you’re sitting beside makes no sense.

I like Bambulabs hardware but I don’t get their obsession with locking down the firmware. Ultimately it’s just a 3d printer that takes an STL and prints it. There is very limited IP in a firmware that needs protection or that couldn’t be figured out by monitoring the I2C or whatever protocol it uses to send instructions to various systems like AMS, camera, printer board etc. Somebody could reverse engineer it already and all this controversy just makes it more likely that someone will.

The ribbon was contentious but most people are familiar with it and it has advantages like taskcentricity and less clutter. LibreOffice has an experimental ribbon that I think should be worked on, mainstreamed and set during installation or in the settings.

UX in other areas should be improved. Lots of little annoyances add up for new users and can break their opinions. It’s not hard to look over the UI and see things which have no business being there, or should only appear in certain contexts, or could be implemented in better ways. I think the project should get some MS Office volunteers into a lab and ask them to do things and observe their problems. I’d have power Word, Excel, Powerpoint users come in and do non-trivial things they normally do and see where they trip up or even if they can do what they need.

Good evidence of astroturfing on Reddit. That Reddit took action and banned the Palantir agents only provides evidence that exposure of the op is the problem. Not evidence that Reddit acts in good faith.

A good question to ask, is what would happen if Lemmy was the victim of astroturfing. It’s decentralized for starters and groups might not even reside in the same place on the fediverse. Also I expect Reddit has monitoring, analytics and tools that could flag behaviour rather than somebody having to go through logs trying to find patterns.

I think Lemmy and other federated platforms have escaped having to deal with these issues simply because someone attempting to astroturf will do it on the biggest platform. So Lemmy escapes not by any technical or administrative virtue but by being smallfry.

An LLM is an ordered series of parameterized / weighted nodes which are fed a bunch of tokens, and millions of calculations later result generates the next token to append and repeat the process. It’s like turning a handle on some complex Babbage-esque machine. LLMs use a tiny bit of randomness (“temperature”) when choosing the next token so the responses are not identical each time.

But it is not thinking. Not even remotely so. It’s a simulacrum. If you want to see this, run ollama with the temperature set to 0 e.g.

ollama run gemma3:4b
>>> /set parameter temperature 0
>>> what is a leaf

You will get the same answer every single time.

I think if I were any non-US government I’d be very seriously thinking about not using Microsoft software at this time, particularly if it connects to the cloud. And that goes for companies with government contracts, or merely companies who are potential targets of industrial espionage.

That said, LibreOffice needs to tap the EU for funding to broaden its features and also improve the UX because it’s not great tbh. It can be extremely frustrating using LibreOffice after using MS Office, in part because the UI is so different, noisy with esoteric actions, and very unrefined compared to its MS counterpart. That needs funding and to get to the point that somebody can pick up LibreOffice for the first time and not be surprised or stuck by the way it behaves.

It’s even worse when AI soaks up some project whose APIs are constantly changing. Try using AI to code against jetty for example and you’ll be weeping.

All AIs are the same. They’re just scraping content from GitHub, stackoverflow etc with a bunch of guardrails slapped on to spew out sentences that conform to their training data but there is no intelligence. They’re super handy for basic code snippets but anyone using them anything remotely complex or nuanced will regret it.

Hardly surprising. Llms aren’t -thinking- they’re just shitting out the next token for any given input of tokens.

My bambu AMS doesn’t work with TPU so I don’t think multimaterial has much benefit in this use case. Basically I have to put the tpu on a spool at the back and disconnect the tubes from the AMS.

I do find multimaterial handy for PLA/PETG although AMS can be annoying if a piece of PLA snaps which often happens.