which is false, but more importantly the tools are entirely different. you can't do the same thing, you have to find the "gimp way" of achieving the same result

I don't like LibreOffice but any actual Open Source dev deserves infinite more respect than those prompstitutes.

I wish the people at my company were like your mother. Instead, they freak out about the smallest of changes in the UI of the same application. God forbid they have to move to another application entirely. Had a user freak out their application wasn't licensed and they couldn't use all functions anymore. Turns out they just managed to somehow hide the toolbar with all the options... These are people anywhere from 20 to 60 years old, btw.

Thanks I’ve added another that does, I thought restrictions on training would include the moment Claude reads the code for modifications too but looking further I think you’re right.

I agree to an extent - a few tools are indeed different, but the majority (including the core ones that 95% of people use) work almost exactly the same. Levels, curves, crop, clone stamp, sharpening, move etc.

Or just slop-machine in short

they made Teams 2, Satan might not get the note because he uses teams.

Not inherently (you can still do things because you want to), but I mentioned wanting to make a living out of selling them. Trivial decompilation makes this completely unviable.

No. If you format aomething in latex it will look like you want. Word and even libre writer will just mess up formatting at the slightest inconvenience

which is false

“Trust me bro, it's completely different.”

Actual apps: tools on the left, layers on the right, the pic in the middle. The important menus are in the same places. Apparently Photoshop users lack pattern recognition.

My experience of using AI is it can save you a LOT of work, but you need to supervise it REALLY closely and give it sufficient code context for it to work properly. It only knows what it knows and if it doesn't know it will make shit up. But if you do use it within its limits it will identify bugs, generate tests, generate a lot of boiler plate and help you be more productive.

Being open source is not a reason to avoid it. However AI does produce slop and unattended it can produce execrable patches which are practically incoherent to anyone used to human coding patterns. So there should be rules about coding style, patch / merge / pull requests that diffs are self evident, concise and obvious. Most AIs will read a guidance file so this stuff can be defined. And if a person uses AI to generate a patch they should review this patch first before daring anyone else to look at it.

Exactly. I also think that the problem was never just implementing stuff, the problem is to support the written code. Lately I stumbled upon a phrase "free as in a free puppy" I think those AI clones are exactly this. You will need to support it with AI, because no one else will be able or will be willing for enough time to do anything there. There is no understanding gained from this and if this bubble ever bursts or at least deflate a bit and prices for a token will get real it will no longer be maintainable.

Yeah, it is not an age thing. I'm very happy that my mother has always been pretty good at programming solving. Although she never did any real tech related job, I believe this is a question of attitude more than anything else.

When she proudly tells me she used my dad's phone to google for a fix for her phone acting up, my inner first year support monkey is proud :)

I disagree.

LibreOffice is a flagship FOSS project - they have delivered genuine benefits to users over a long period of time with no shady financing arrangements or other encumberances - few projects can claim this outstanding status.

They could not have achieved that status if they were dancing along to Microsoft's merry jig. With very limited resources LO can't implement microsoft's proprietary features.

OnlyOffice by comparisson is opensource but has a paid tier intended for corporate clients. That being the case they really need to offer better compatibility.

Tinfoil hat time:

A year ago, I did some reverse engineering. I figured a pattern matching engine has to be good matching patterns, so I threw the messed-up output of Ghidra into Claude to clean it up and to give me an overview of what the fuck happens in a certain part.

I was absolutely surprised. Sure, it got some things wrong, but it knew way more than it should have been able to guess from that short bit of fucked up, optimised not-quite-C. Including guessing context that it should not be able to guess. This was a firmware with not-yet-released functions that, to my knowledge, shouldn't exist outside of this bit of binary firmware. At least not for us non-NDA mortals.

At the time, Mistral had a new-ish model as well, so I figured I'd give it a try. It worked as expected, cleaned up things, gave variables reasonable names, made educated guesses and so on, but didn't act as if it had background info. And I was using Sonnet 4.x, so the small Claude which shouldn't be that much larger.

Since then, I have the conspiracy theory that the Claude people threw a bunch of proprietary sources in the training, too. I have no idea how, maybe aggressively training on user inputs? Anyway, this does not help to allalleviate my suspicions.

On the topic it means AI has its use cases. Outright banning it is not the solution. The actual Problems are PRs that the authors dont check/understand.

They also vibe coded recreations of apps from Adobe suite, MS Office and more. It's very fresh news and "development" is moving at rapid pace, so it wasn't hard to miss. For example first commit of WordCraft was created six days ago and now they are already distributing version 0.5.0.

I'm sure it's all but officially confirmed, that all the people upload to it goes to training data unsanitised. And Microslop for sure uploaded entirety of their codebase to all the bots. This is at least part of the reason why claude is so surprisingly OK at reverse-engineering all the big stuff.

Since then, I have the conspiracy theory that the Claude people threw a bunch of proprietary sources in the training, too.

An enormous amount of people use this technology for privat company data without batting an eye; not just software developers, but business people for reports, etc.

They all throw their confidential data in there without even considering the implications, just "trusting" Anthropic. claude code bots are running on tens of thousands of proprietary code bases that used to be private, and of course they are using all this data to train their model.

The person behind all these *craft projects has (or claims to have) made millions from his AI image / video creation services and scam-enabling deepfake voice cloner. The projects are linked to and funded by his commercial endeavors. He's also extremely exuberant about real developers and creators losing their jobs, and suffers from massive AI psychosis. Unfortunately I don't think he or this project is going anywhere.

I don't think any of it will ever actually work, but he's very good at making outlandish comments and drumming up hype, so we're going to keep hearing about it anyway.

Then people are surprised that claude code gets better at coding

a couple of months ago i asked about this one of local Forbes editors and they explained this phenomenon as "because it is expected to code by hand but managing to vibe anything is amazing". i decided not to continue this conversation because that's fucking stupid

Judging from their past performances and ideas, it was already vibe coded long before AI. Based on the vibes of whoever was in charge at that point.

OnlyOffice is "largely FOSS but with issues".

Better at replicating closed source software.

It surely gets better at coding in general, after all there are smart people at anthropic with near limitless budgets, but my conspiracy thesis is that it gets a lot better at replicating stuff it has seen before, but really shouldn't have.

People aren't as opposed to paying as the loud people on the internet would suggest. Particularly if it is the best software around, hence why Photoshop, AutoDesk, etc have such a stranglehold. Even more so if you're marketing to a business. Is a company really going to use a vibe coded app, or pay their engineer to spend time decompiling and reverse engineering your app while production grinds to a halt, and THEN somehow maintain vibe code, or are they just going to pay for the thing that works and that they can use right now? Similarly, game piracy has never been easier, but Steam makes money hand over fist. Many piracy sites I've seen refuse to crack indie games.

I'm not telling you what to do with your life though, it's a gamble to be sure.

"In favor" is a strong phrase. "Understanding the inevitable" might be more apt.

What barely works right now will get better. What was slop will surpass the best humans can code soon. I am old enough to remember when "Made in Taiwan" and "Made in Japan" were racist pejoratives for bad quality.

It's great to love your horse and your ferrier business, but when you're watching Model T's roll off the assembly line, even with initial build quality issues, you can see the world is changing.

They claim it's "clean room", so no decompilation involved. I'm just waiting for the lawsuits to arrive, so we may learn how much of that is true. The big ones probably are already checking the code.

I don't think it's a apt description, this so the first time human knowledge has in anyway been "automated".

It's either "doesn't work" or "does work".

And once it truly does work it'll be a different store. These sloperators thing that they have AGI like if they use the correct magic words a real enterprise app will come out. But it's not true, they are riddles with technical debt, drag, security vulnerabilities and bugs. No matter how many times you sift through them with the LLM.

Once AGI has arrived its a different story. Once coding is "solved" via a AI it's pretty much the singularity. software development requires critical thinking skills that once acquired will allow them to pretty much do any knowledge job on the planet and actually self improve.

LLMs aren't it, and when something real comes along I'll be ok board. LLMs have their use cases but anything outside of automating a call center or a voice assistant is pretty much a anti-pattern.

These aren't model Ts these are horse drawn wagons with 500 horses so they say they can have as much horse power as a real car does.

they're desperate for ai to be profitable.

you're right

That's a bit of a risky thing to do. They need to be careful not to introduce some proprietary code.

There's Photogimp

Damn; looks like grandma is going to outlive us all

True, however iirc the tools panel layout can be adjusted in standard GIMP — maybe not to the width of one icon, and by default the panel is also used for the tools options. Afaiu there was more sense in Photogimp when GIMP had floating windows, those really were dark times.

My main issue with GIMP is that I can never memorize the location of the buttons for less-frequently-used tools, and the icons are not too recognizable. But that can be fixed with an icon theme, and perhaps with reordering the buttons if that's indeed possible.

I specifically asked our CEO about this just to cover my ass. It seems insane to trust these guys at all. Here in the USA, small businesses have de facto no legal protection from large businesses.

Even in undergrad, I swear using good LyX templates nets you half a grade just for beautiful presentation. And if you throw Zotero in there for the automated citation management, it's glorious for research papers.

There is another twist to this in that the OOXML format didn't actually define the bugs in the format and MS Office itself didn't actually follow the officially defined specification they pushed through. This inevitably led to every product that tried to make use of the official spec (e.g. Open Office) failing to be compatible with MS Office. It took the open source devs many years to properly duplicate the buggy mess that is the real MS Office formats.

Absolutely. Especially if you have company info / IP that someone else would pay a lot for. This is a question for lawyers and managers to negotiate with the lawyers at OpenAI, Anthropic etc, not for us little people to decide. I'm not throwing company data in an online service that is not cleared for us to use, and I don't even work with spicy stuff. Back in my IT days we would have called that a data breach, which would have lead to something from an unpleasant hour in the boss' office to a promotion to customer.

Providers like MS at least should have figured out that the hypothetical ability to take your data "to go" is important, but nothing they need to worry about being actually used too much, especially in businesses.

One is the anxiety over screwing up and losing your data when you solely manage it. So huge pressure to use "a" reputable company as a steward of the data.

Another is that the user and the spend are disconnected in business scenarios. If attaching a 'cloud' document to something in the app is two clicks, but 4 clicks to do the same from desktop, then the users will want to just stick to the cloud to save them a few clicks at least. They will demand their IT provide microsoft solutions and the IT will shrug and say "well, everyone gets microsoft solutions". Back in the day my company tried to split the difference by purchasing perpetual licenses only for every third edition. Now Microsoft has made that just not viable so you either just pay for as long as you use it or you go elsewhere.

It's also why I don't think any of the providers are actually trying to screw up desktop component prices for fear of local models ruining their day. The tech market has shown they actively jump at the chance to surrender their flexibility to providers, no need to worry about them suddenly deciding to in-house things.

Meta had an explicit higher price for using their model without retention of prompts for training. This all but confirms your theory.

In terms of whether the 'clean room' approach is as critical in Europe or not, I can't be sure. It might be more permissable for someone to use a decompile as a reference and also directly work on a replacement. However I'm pretty sure starting with a decompile of proprietary software as a direct codebase would run afoul of issues.

Otherwise, imagine you have an HEVC encoded rip of a 4K blu ray. You transcode it to AV1. The bytes are totally different, and even the resultant frames are likely not even pixel for pixel identical. I don't think copyright law would deem that not a violation, regardless of jurisdiction.

It's standard operating practice to steal everything. The world's full of dumb fucks.

Wait... wait... you believe Facebook?

Long live Libre devs!! Been with yall for over a decade

It's not really stupid. Things that are unusual are more noticeable, and garner more comment. We've grown used to devs tirelessly doing what they do, to the point most people don't think about where their software comes from, but a machine sort of doing it (poorly, so very poorly) is novel, so people talk about it.

Currently, software mostly working as it should is the norm, so people don't talk about it much. When the wave of slopcoded disasters starts to grow, people will talk about it, because software is suddenly not working as it should.

and yet no one is celebrating a piece of software well done

Makes me wonder how much of the frontier models hacking is just "our devs uploaded the codebase to Claude and suddenly it knows where all the backdoors are"?

Yes, personally I don't have a huge problem if someone uses an LLM, that's been trained only on opensource code, on a project, but that someone has to put their name to the code, documentation, tests or anything elsr produced, and take full responsibility for it. As in, they suffer the same consequences if it causes an issue as if they'd hand coded it, with particular attention paid to whether the result is maintainable, and whether it breaches any licences.

Of course, most LLMs have, by now, injested a lot of proprietary code that people have mindlessly sent in queries, rendering all of their output tainted.

Oh, facebook will steal the data even if they say they won't. What's interesting is they put a price on blatant prompt copying.