I created all the designs for flyers, announcements, invitations etc. in Affinity for my wife. I migrated to Linux and didn't like any of the apps. I was lost.
Then Affinity announced they'll add Linux support.
But for some time now my wife just uses Codex/Claude to do her day-to-day nothing special design work.
I feel, just like ArtCraft, Gimp will not be relevant in some years (months?).
There are people who want to use Gimp (or Affinity etc), and there are people who want to have used Gimp. It’ll only lose relevance for the second group.
AI detectors are notoriously unreliable. As someone who's read too much slop in the past few years, this doesn't strike me as AI. There's typos like "coult" instead of "could", and clunky sentences like this:
> Furthermore such legacy tests directly hijacking core code come with their own problems, such as breaking all too often (and not because they detected a bug), all while not being able to detect whole categories of bugs.
That’s not really a clunky sentence. While it might not be Pulitzer-worthy, it clearly conveys a set of interconnected issues. Arbitrarily splitting it into a few sentences wouldn’t make it better.
Pangram is just routinely spammed in every other HN thread, but as far as I can tell, it's not much more reliable than rand()%2, and that's only for cases where the "author" doesn't even care to hide the text was generated.
It doesn't and cannot work in general case. It's trivial to make LLM write in any style, or have it prorduce arbitrary number of typos and grammar mistakes. Most people just don't bother to add an extra line to the prompt.
Several independent audits (not counting Pangram's own research, of course) all finding extremely low or zero false positive rates and pretty good recall:
Also, the fact that when a public bounty was placed on a known-human text of >1000 words that Pangram falsely flags as having a >75% AI content, the only example anyone found was this one:
https://twitter.com/Nition/status/2122500484007395664
> Pangram is just routinely spammed in every other HN thread, but as far as I can tell, it's not much more reliable than rand()%2
As far as you can tell... based on what?
> It doesn't and cannot work in general case. It's trivial to make LLM write in any style[...]
If you asked me two years ago whether AI detection was possible, I'd have answered no on these grounds. It turns out, however, that RLHF as it's currently done causes LLMs to collapse to very specific styles, and labs are either unwilling or unable to figure out a way to fix it, which is why both humans and models are in fact able to detect AI-written text. Base models don't have this style collapse and indeed aren't detected by Pangram, but very few people are willing to use base models. Similarly, finetuning a model to write in a specific style would probably suffice to evade detection, but very few people are willing to put in this much effort. The result is that it's hard to be sure that any given text isn't AI-written, but most AI-written text is extremely blatant.
It doesn't work now because people have brain rot and talk like AI. It only worked because they mapped a couple tics and then pumped it across the nerdfluencer sphere.
I think the text was originally written entirely by AI (the structure is typical). Then it was partially rewritten.
Some clues that caught my attention on their own:
“Each test launches a real instance of GIMP and performs a small task. Then it reads the result from GIMP. If the result is incorrect, the test fails and tells us what went wrong.”
That’s classic Claude; just like what I read every day. His approach to humanizing everything
And Pangram IS conservative in its estimates of results. With a text this long, I have no doubt about its accuracy.
Where did you get that quote from? It definitely looks like an LLM rewrite of a human-written paragraph from the post
> Each test starts a real GIMP and does one small task. Then it reads the result from GIMP. If the result is not correct, the test fails and tells us what went wrong.
Huh? It does appear both in the Pangram check and in the blog, and in the first commit in the history too. Right after the "What the tests do" heading.
That's not how it works. A 40% score means Pangram detects 40% of this text as AI (and the rest as human) at its usual confidence, rather than assigning a 40% probability to it being AI.
Or maybe don't treat pangram as some sort of magic oracle on AI generated writing. Most recently they gave out a prize because their detector scored a poem from 2006 with the phase "the night we ate whale" repeated four thousand times, as 100% AI generated.
I don't think you draw the same conclusions from this search as me. Multiple people searched for ~a week with a bounty on the line and the only long false positive they could find was this poem. Meanwhile, if Pangram wasn't extremely sure of their accuracy, they could simply, you know, not participate in the bounty. Moreover, the person who submitted that winning entry does not consider it to be in the spirit of the rules, see e.g. this thread.
At the risk of being obvious, Pangram has heavy incentive to give you something to say about the Pangrammyness of text you plug into Pangram.
If it mostly said "nope! No AI here!" you'd walk away and wouldn't Pangram your comment with mentions of Pangram, which would be a missed opportunity for you to give free brandgram awareness to Pangram.
Pangram is currently the most reliable tool for avoiding the flood of low-quality AI content on the internet.
There are independent studies on its effectiveness.
I used it because I had strong suspicions about certain passages in this text. Otherwise, I would never have used my Pangram credits on this.
Anyone who regularly reads AI-generated text will quickly recognize the few AI quirks in the text.
At the risk of being obvious, you can easily find extensive research or check for yourself than Pangram does consistently produce "100% human" scores for human writing. The whole reason why it gets a lot of "free brandgram awareness" is that it's not like the dozen shitty detectors out there - what's the last time you've heard about GPTZero?
Does GIMP have a very restrictive AI policy or something? I get not wanting to ship a gigantic pile of UI tests without thorough review, but basically any model released in the last six months would be able to write UI tests covering every single control in an afternoon.
Yeah, i'm sure they want to burn their own tokens, deny others to use AI, to hog all that foss money for themselves. Definitely not related to not wanting to review 10k ai generated prs no siree
It's ok to have principles. Some folks have extremely valid reasons to hold a hard no stance on AI. It's not me, but I respect that position as principled, logical, and reasoned. Just a very different set of values from me.
I liken it to my vegetarianism. I'm not morally against eating animals. I have no religious convictions about it either. I just look at factory farming and think "that seems like torturing animals at scale to me" and choose to not participate.
The people looking at AI and taking a "none for me thanks" stance are looking at externalities and saying they don't like them, and they dont like them enough to not participate.
But there’s a dozen other invisible tradeoffs. It’s not a simple binary. The Amish don’t have their own grapes year round, for instance. The Amish would need to dedicate a major percentage of their population to farming, given the drastically lower productivity.
Our food supply has perfectly healthy options. People choose to buy and eat junk, then blow up to 400lbs. You can gild the lily by arguing over organic vs Amish-grown in terms of micronutrients and traces of pesticides, but even an inorganic ear of corn is 10x healthier than what the average American eats.
what's bizarre about it? if all you care about is things appearing on the screen then vibecode away. but some people also have strict requirements for the architecture/quality/performance of the code underneath.
Your response is essentially a strawman. Use of an LLM doesn't need to be "vibe coding", and LLMs are perfectly capable of producing cromulent, performant, well architected code in the right conditions and in the hands of a developer who knows how to use it.
Personally I consider it bizarre because the entire thesis and history of computing, going back to the term "computer" referring to a human who did tedious arithmetic by hand, was leveraging systems to accomplish a goal. Except now. Because now for (often) vague purity reasons, we should no longer reach for that lever.
That "vague purity reason" is almost always software developers so engrossed in their art form that they are more interested in the code than the product. There's not necessarily anything wrong with that, but if your goal is to ship a product, choosing not to reach for a lever to make the product better meets my definition of "bizarre".
It's bizarre in the same sense that abstract art is be bizarre. It can be interesting, novel, and even thought provoking; but often lacks functionality or utility to make it applicable to the lives of people. That's what non-LLM projects going forward will be.
If you start with the assumption that AI is the devil, than rejecting the devil and all the fruit he offers is logical.
If you start with the assumption that AI is a tool, then finding uses for the tool is logical.
Seems to me that some people really love riding a purity spiral, and collectively they have chosen a new tool to fill the role of the impure thing to be cast out.
Someone can believe AI was trained on material in a way that violated the consent of authors and artists and musicians. Someone can see the value of the tool, but not want that value to be primarily captured by megacorps. One can believe that the tool is having a deleterious effect on students, workers, and artists of all kinds, as it's being deployed. One can't be concerned about the added environmental effects, the impact to world markets, or the shortages it causes to consumer hardware. One can be worried about the copyright implications (can AI produce copyrighted material?) and not want to risk losing their IP. One can believe that art is necessarily a human enterprise, and art produced mechanically is just images and sounds, but not art nor music.
None of these are "the tool is the devil". All of these are "the tool has externalities, and those externalities come with a cost I am not comfortable paying".
I'm a vegetarian. Not because I think eating meat is immoral, but because I believe factory farming causes undue suffering to animals and the people who raise and process those animals. So I make a choice for myself, in my own life, and don't moralize to others.
It's the same thing. Some folks are making choices for their own lives based on their values for those externalities. You value energy use differently than they do. You value art differently than they do. You value corporate control over communications differently than they do. Neither of you is right, but standing here and pretending an eminently justifiable difference of opinion is just purity signaling is boorish.
Do you think its resonnable that the Gimp project are outputting handfull of human written tests when there is a good chance that they could use a tool to write thousands of tests in an afternoon?
They may be proud of not paying the externalities that the evil megacorps are demanding, but that warm blanket of purity is no protection from being outcompeted.
I think it's a real shame that so many people have set themselves against an emerging technology for reasons that on closer analysis seem like just another round of prejudice and superstition.
From what I've seen, it's less spiritualism and more "they're taking our jobs!" kind of fearmongering. Like all "taking our jobs" fears, whether it's logical or not depends purely on whether they're actually taking these jobs or not.
> You have to serve users, not hold code up as a sacred relic.
They are serving users, by not allowing AI slop. Software developed by people and not an LLM is of much higher quality, and I'm glad they are taking a stand.
The Free Software movement itself is holding up principles as a relic. It has both helped, and hindered the movement in the past. I see no reason for them to stop.
Someone else is welcome to fork GIMP and use AI to serve the users, if they so like.
At least DNA has a reward signal to clean up tech dept. Everything is use-it-or-loose-it. Don't use your eyes because it's always dark? Bye bye. Claude would maintain that as a fallback for all eternity.
It wasn't vibe coded, it evolved over millions of years. And we ended up with our eating hole right next to our breathing hole because there's no way up back out of that "decision".
Does free open source software “have to” serve users? You clearly aren’t the user they’re trying to serve. That’s okay though; there’s something for everyone.
It’s frustrating to create tests with models, not sure if it’s everyone’s experience, but every time I asked it to create tests, it makes the tests conform to the sometimes broken code. Rather than calling out what the logic/code is wrong, it will make the tests conform to whatever is already broken.
Prodding an agent vs doing interesting things with my time? What should I choose?
To be fair, some people like to prod. But, for me, I'd rather take hours reading manuals than do trial and error prodding for 30 minutes (even if I knew it would be that short). Yes it's not optimal in terms of time, but it's better for my soul and well being.
> Yes it's not optimal in terms of time, but it's better for my soul and well being.
It is also optimal in terms of time in the long run. The prodder will have to prod over and over again because he has learned nothing. You on the other hand will actually understand what is going on and will be able to either do it yourself or give precise instructions to the machine next time. Don't let yourself become a meat proxy.
It's a good use of your time to ask the agent to reflect and update it's skills if you intervene to manually steer it. Why wouldn't you want it to learn from your guidance?
The problem I've had is even when it can explain back the issue, it still writes for itself awful instructions with extra fluff, strange points of focus, and borderline misinterpretations of what I said.
Maybe that's not a big deal, they can still be generally useful, but they seem important to commit with the project if they will be used going forward. I hate the idea of a massive amount of slop riding along with the code, and slop begets slop.
The idea I'm chasing now is writing the documents myself with the hope that enough examples and a document about writing documents will be enough to allow it to write clearly.
It can’t magically know expectations. If the code is well documented or there’s decent associated documentation, I’ve found it has no issue writing tests.
Tests are usually the best documentation and often the only documentation. If humans bother to write anything it should be the tests, not ambiguous human language...beyond the high level 1-liner overviews.
True. It's tough useful to build non-regression tests, that is build tests that test the current behavior, assuming it correct, to possibly spot bugs in the future.
Not using AI to at least writes tests in this day and age is crazy work. I started a desktop application 2 months ago and it has almost 3,000 tests now, MANY of them UI tests.
AI is a bit too excellent at writing tests. The number of tests doesn't really convey any information on test coverage. At least I've found that agents that naively improve codebases will write tests for pretty much anything, including irrelevant behavior that has little value in ensuring stability, security or correct behavior. It would probably help to run a test review stage with a strong model, where agents attempt to correct and consolidate tests to guard behavior rather than ensuring that theoretical code paths are followed.
In my experience, designing tests cases to ensure that they're othogonal and verify your invariants (domain or technical) is very mentally taxing and not something you can truly delegate. The actual typing maybe, but testing is a lot of duplicate code and you can go far with just copy-pasting stuff around and building helpers.
Using AI to write tests in this day and age is crazy work. The machine is great at writing tests, but the quality of the tests is awful. It will write redundant tests, it will write tests that assert too many thing in one, it will not write tests that need to be tested. It's the part where I have to clean up the most every time. Quantity is not quality, and you really don't want to skip on the quality of your quality assurance. I'm writing a React + Django full-stack application, so there should be more than enough training data out there.
Funny how GIMP spawned the Gimp Toolkit, aka GTK, which is widely used in Linux, and yet GIMP's own UI seems to have long languished as a rather outdated design. Glad that they have tests, though, and hope they continue to improve it.
They made a blog post to announce that they’d written thirteen unit tests? Sure, there’s value in handmade software and all that but GIMP is not really a particularly excellent piece of software - if anything was to benefit from unleashing the agents it would be GIMP.
It'd be really interesting to look back the comparison between projects with strict non-AI policies (e.g. GIMP) and ArtCraft in 2027.
My personal prediction is that ArtCraft will blow all the old non-AI projects in similar fields out of the water and GIMP will be remembered as a thing of the past. But if I got proven wrong I'd be happy too: at least we will know the value of non-AI policies.
GIMP is far from the best FOSS image editor for many use cases. IMO for many simple editing tasks that people usually need unless they're professional photo editors, even Inkscape, a primarily vector editor, is a better photo editor than GIMP.
Inkscape is what made me believe ArtCraft would be the clear winner, btw. It's very hard to imagine using Inkscape ever again since I tried VectorCraft. (And I'm fully aware that some Path Effects can only be done quickly in Inkscape.)
The blog article doesn't say they have written thirteen unit tests.
> *GIMP already has unit tests*. A unit test checks one small part of the code, for example how GIMP reads an XCF file. But our unit tests were not opening a real window. So it coult not tell us if a dialog opens, or if a menu item did its job.
> Note that GIMP *already had tests of GUI code*, usually interacting directly with the core code, and sometimes even simulating input devices’ events (pointer movements, clicks, drags…). But nothing can beat interacting with the real interface. Furthermore such legacy tests directly hijacking core code come with their own problems, such as breaking all too often (and not because they detected a bug), all while not being able to detect whole categories of bugs.
> UI tests check the program as a whole, in the same way as you use it.
Lucky you. Making things testable seemed hard to me quite often and ensuring that the tests covered all the difficult bits without effectively mocking it to death so that one was testing almost nothing.
Timely, as I removed Gimp from my Macbook today after successfully testing PhotoCraft [1] for a week (the AI Photoshop clone written in open-source rust which has had an absured number of updates since you last tried it).
It makes me so HAPPY to remove the dinosaur Gimp, and to run "Adobe" software for free (it's good enough for me, maybe good enough for you). I expect to lose Internet points for saying that.
I have a long history with Gimp. I created Gimpshop... the Gimp maintainers did not appreciate that little stunt, I'll tell ya! :D
I have canceled my Adobe subscription as of last month and have fully moved over to Photocraft. Even if development of Photocraft stopped today, I would be able to do whatever I need to get done in Photoshop without paying Adobe or wrestling with Gimp. It's good enough for me as well. I had been using Adobe Photoshop for more than 30 years now and was bought in enough to spend a significant amount of effort making Gimp act like Photoshop. Bye Photoshop, bye Gimp, bye Gimpshop.
Same with VectorCraft. I spent over a decade looking for an Illustrator alternative. Muscle memory and specific workflows won out. VectorCraft is extremely close.
AI bros can't stand the thought that not everyone agrees with them about LLMs. It's like a religion to them, they feel compelled to get everyone else on board. Never mind that if they are right, it will become apparent to everyone without them having to preach.
When a group of people relentlessly insist on hijacking any discussion to refocus the topic onto something unrelated but more interesting to them it stops becoming a preference and starts looking more like bullying at the hands of a religion or cult. For instance:
So why do you need to pivot every discussion about "software development process" to AI? Do you believe that the GIMP authors have not yet heard the good word? Do you think they need to be saved from their primitive non-AI practices?
There's a difference between discussion and proselytizing. And, yes, that rises to the level of bullying when AI bros relentlessly drag every discussion back to "WHY U NO AI???". It's absolutely bullying when you make it impossible to discuss non-AI related subjects.
The tests should simply take a screenshot and send it to an LLM with the question "Is the UI absolutely horrible?". If the answer is yes then the UI is working as expected.
Tests are a good first step, here's hoping it'll get a UI next :)
I created all the designs for flyers, announcements, invitations etc. in Affinity for my wife. I migrated to Linux and didn't like any of the apps. I was lost.
Then Affinity announced they'll add Linux support.
But for some time now my wife just uses Codex/Claude to do her day-to-day nothing special design work.
I feel, just like ArtCraft, Gimp will not be relevant in some years (months?).
For some people GIMP is irrelevant for the last ~25 years. I was using it in 2000, I'm still using it in 2026.
It's a matter of getting used to things. I feel lost in Affinity tools, for example, despite having a license, I use them less and less.
Use this: https://getartcraft.com/ It is open source, built with Rust and very fast and works great and is very similar to Adobe Photoshop
Tell me you didn't read the whole comment you replied to without telling me...
You switched to Linux, but you don't want to use Linux things. Why not just stick with Windows (or whatever you were using before)?
My son learned GIMP in school and refuses to learn anything else.
There are people who want to use Gimp (or Affinity etc), and there are people who want to have used Gimp. It’ll only lose relevance for the second group.
> So it coult not tell us if a dialog opens,
It feels good when I see spelling mistakes. It feels organic, human-sourced, non-AI. It feels like a good old village again.
BTW I love GIMP for everything it did to the world of graphics work. Such an awesome thing. Thank you folks.
[flagged]
Pangram detects AI in the text at +40%. That's a bit disappointing coming from Gimp, which prohibits the use of AI.
It seems hypocritical and raises questions about whether the lead maintainers are following their own guidelines.
edit : typo
https://www.pangram.com/history/be1f7073-5b2c-4431-9c5f-dae7...
AI detectors are notoriously unreliable. As someone who's read too much slop in the past few years, this doesn't strike me as AI. There's typos like "coult" instead of "could", and clunky sentences like this:
> Furthermore such legacy tests directly hijacking core code come with their own problems, such as breaking all too often (and not because they detected a bug), all while not being able to detect whole categories of bugs.
That’s not really a clunky sentence. While it might not be Pulitzer-worthy, it clearly conveys a set of interconnected issues. Arbitrarily splitting it into a few sentences wouldn’t make it better.
It is definitely a clunky sentence, and doesn't read much like a claude commmit message
Old AI detectors are notoriously unreliable, but not Pangram, which has an extensive record of reliability.
What record? Where is it? What is it based on?
Pangram is just routinely spammed in every other HN thread, but as far as I can tell, it's not much more reliable than rand()%2, and that's only for cases where the "author" doesn't even care to hide the text was generated.
It doesn't and cannot work in general case. It's trivial to make LLM write in any style, or have it prorduce arbitrary number of typos and grammar mistakes. Most people just don't bother to add an extra line to the prompt.
> What record? Where is it? What is it based on?
Several independent audits (not counting Pangram's own research, of course) all finding extremely low or zero false positive rates and pretty good recall:
https://epoch.ai/data-insights/ai-detectors-false-negatives
https://arxiv.org/html/2608.11256
https://link.springer.com/article/10.1007/s40979-026-00226-w
Also, the fact that when a public bounty was placed on a known-human text of >1000 words that Pangram falsely flags as having a >75% AI content, the only example anyone found was this one: https://twitter.com/Nition/status/2122500484007395664
> Pangram is just routinely spammed in every other HN thread, but as far as I can tell, it's not much more reliable than rand()%2
As far as you can tell... based on what?
> It doesn't and cannot work in general case. It's trivial to make LLM write in any style[...]
If you asked me two years ago whether AI detection was possible, I'd have answered no on these grounds. It turns out, however, that RLHF as it's currently done causes LLMs to collapse to very specific styles, and labs are either unwilling or unable to figure out a way to fix it, which is why both humans and models are in fact able to detect AI-written text. Base models don't have this style collapse and indeed aren't detected by Pangram, but very few people are willing to use base models. Similarly, finetuning a model to write in a specific style would probably suffice to evade detection, but very few people are willing to put in this much effort. The result is that it's hard to be sure that any given text isn't AI-written, but most AI-written text is extremely blatant.
It doesn't work now because people have brain rot and talk like AI. It only worked because they mapped a couple tics and then pumped it across the nerdfluencer sphere.
Why don't you let me fix you some of this Mococoa drink? All natural cocoa beans from the upper slopes of Mount Nicaragua. No artificial sweeteners.
"The article detects AI in the text at +40%"
Huh?
I don't know how, but my iOS keyboard has completely replaced “Pangram” with something else...
Yeah, that's called autocorrect
All the articles on HN that I suspected as AI-generated got ~100%. So a 40% means likely no AI in my dictionary.
I think the text was originally written entirely by AI (the structure is typical). Then it was partially rewritten.
Some clues that caught my attention on their own:
“Each test launches a real instance of GIMP and performs a small task. Then it reads the result from GIMP. If the result is incorrect, the test fails and tells us what went wrong.”
That’s classic Claude; just like what I read every day. His approach to humanizing everything
And Pangram IS conservative in its estimates of results. With a text this long, I have no doubt about its accuracy.
Where did you get that quote from? It definitely looks like an LLM rewrite of a human-written paragraph from the post
> Each test starts a real GIMP and does one small task. Then it reads the result from GIMP. If the result is not correct, the test fails and tells us what went wrong.
Your quote doesn't appear in your Pangram link or in the commit history for the blog https://gitlab.gnome.org/Infrastructure/gimp-web/-/commits/t...
Huh? It does appear both in the Pangram check and in the blog, and in the first commit in the history too. Right after the "What the tests do" heading.
That's not how it works. A 40% score means Pangram detects 40% of this text as AI (and the rest as human) at its usual confidence, rather than assigning a 40% probability to it being AI.
Or maybe don't treat pangram as some sort of magic oracle on AI generated writing. Most recently they gave out a prize because their detector scored a poem from 2006 with the phase "the night we ate whale" repeated four thousand times, as 100% AI generated.
https://x.com/max_spero_/status/2122032840204226687
I don't think you draw the same conclusions from this search as me. Multiple people searched for ~a week with a bounty on the line and the only long false positive they could find was this poem. Meanwhile, if Pangram wasn't extremely sure of their accuracy, they could simply, you know, not participate in the bounty. Moreover, the person who submitted that winning entry does not consider it to be in the spirit of the rules, see e.g. this thread.
https://twitter.com/Nition/status/2122500484007395664
[flagged]
At the risk of being obvious, Pangram has heavy incentive to give you something to say about the Pangrammyness of text you plug into Pangram.
If it mostly said "nope! No AI here!" you'd walk away and wouldn't Pangram your comment with mentions of Pangram, which would be a missed opportunity for you to give free brandgram awareness to Pangram.
Pangram!
Pangram is currently the most reliable tool for avoiding the flood of low-quality AI content on the internet. There are independent studies on its effectiveness.
I used it because I had strong suspicions about certain passages in this text. Otherwise, I would never have used my Pangram credits on this.
Anyone who regularly reads AI-generated text will quickly recognize the few AI quirks in the text.
Humans can do better. They DO better
Thank you for your Pangram primer, I had no idea!
Pangram!
At the risk of being obvious, you can easily find extensive research or check for yourself than Pangram does consistently produce "100% human" scores for human writing. The whole reason why it gets a lot of "free brandgram awareness" is that it's not like the dozen shitty detectors out there - what's the last time you've heard about GPTZero?
[flagged]
[flagged]
[flagged]
Please regale us with your tales of things that should or deserve to continue to exist?
Everyone can just vibe his own bespoke image editor now, programming is solved, everything is open source, all hail the machine god /s
Not yet but would be nice.
> We have 13 tests at this time.
Does GIMP have a very restrictive AI policy or something? I get not wanting to ship a gigantic pile of UI tests without thorough review, but basically any model released in the last six months would be able to write UI tests covering every single control in an afternoon.
Turns out you're right! No AI allowed https://gitlab.gnome.org/GNOME/gimp/-/blob/master/.gitlab/me...
> for contributors without Developer role
From the blame PR:
> Of course, core developers don't need to check this monotomous checklist. This is only useful for the newcomers.
So this is saying core devs already know not to use AI, only new developers shouldn't.
It's quaint at this point. Embrace the future.
The future seems to be blocking external contributors and running the equivalent AI agents yourself.
Yeah, i'm sure they want to burn their own tokens, deny others to use AI, to hog all that foss money for themselves. Definitely not related to not wanting to review 10k ai generated prs no siree
Huh, that's fascinating. I wonder if this policy is also meant to apply to the GIMP blog - if so, this very article is probably a violation.
Meanwhile I'm watching a community of hundreds of agents reworking game engines and chat apps. And it's progressing fast.
You have to serve users, not hold code up as a sacred relic.
Our own DNA was vibe coded. It's a mess. Stochastic hill climbing can get pretty far, and it can reach surprising fitness.
It's ok to have principles. Some folks have extremely valid reasons to hold a hard no stance on AI. It's not me, but I respect that position as principled, logical, and reasoned. Just a very different set of values from me.
At this point it's like the Amish. Sure I can respect their principles, but at the same time I find it bloody bizarre.
I liken it to my vegetarianism. I'm not morally against eating animals. I have no religious convictions about it either. I just look at factory farming and think "that seems like torturing animals at scale to me" and choose to not participate.
The people looking at AI and taking a "none for me thanks" stance are looking at externalities and saying they don't like them, and they dont like them enough to not participate.
But there's no denying Amish food is much healthier and higher quality than regular store bought products (with some exceptions I guess).
But there’s a dozen other invisible tradeoffs. It’s not a simple binary. The Amish don’t have their own grapes year round, for instance. The Amish would need to dedicate a major percentage of their population to farming, given the drastically lower productivity.
Our food supply has perfectly healthy options. People choose to buy and eat junk, then blow up to 400lbs. You can gild the lily by arguing over organic vs Amish-grown in terms of micronutrients and traces of pesticides, but even an inorganic ear of corn is 10x healthier than what the average American eats.
what's bizarre about it? if all you care about is things appearing on the screen then vibecode away. but some people also have strict requirements for the architecture/quality/performance of the code underneath.
Your response is essentially a strawman. Use of an LLM doesn't need to be "vibe coding", and LLMs are perfectly capable of producing cromulent, performant, well architected code in the right conditions and in the hands of a developer who knows how to use it.
Personally I consider it bizarre because the entire thesis and history of computing, going back to the term "computer" referring to a human who did tedious arithmetic by hand, was leveraging systems to accomplish a goal. Except now. Because now for (often) vague purity reasons, we should no longer reach for that lever.
That "vague purity reason" is almost always software developers so engrossed in their art form that they are more interested in the code than the product. There's not necessarily anything wrong with that, but if your goal is to ship a product, choosing not to reach for a lever to make the product better meets my definition of "bizarre".
It's bizarre in the same sense that abstract art is be bizarre. It can be interesting, novel, and even thought provoking; but often lacks functionality or utility to make it applicable to the lives of people. That's what non-LLM projects going forward will be.
If you start with the assumption that AI is the devil, than rejecting the devil and all the fruit he offers is logical.
If you start with the assumption that AI is a tool, then finding uses for the tool is logical.
Seems to me that some people really love riding a purity spiral, and collectively they have chosen a new tool to fill the role of the impure thing to be cast out.
I think you aren't being thoughtful enough.
Someone can believe AI was trained on material in a way that violated the consent of authors and artists and musicians. Someone can see the value of the tool, but not want that value to be primarily captured by megacorps. One can believe that the tool is having a deleterious effect on students, workers, and artists of all kinds, as it's being deployed. One can't be concerned about the added environmental effects, the impact to world markets, or the shortages it causes to consumer hardware. One can be worried about the copyright implications (can AI produce copyrighted material?) and not want to risk losing their IP. One can believe that art is necessarily a human enterprise, and art produced mechanically is just images and sounds, but not art nor music.
None of these are "the tool is the devil". All of these are "the tool has externalities, and those externalities come with a cost I am not comfortable paying".
I'm a vegetarian. Not because I think eating meat is immoral, but because I believe factory farming causes undue suffering to animals and the people who raise and process those animals. So I make a choice for myself, in my own life, and don't moralize to others.
It's the same thing. Some folks are making choices for their own lives based on their values for those externalities. You value energy use differently than they do. You value art differently than they do. You value corporate control over communications differently than they do. Neither of you is right, but standing here and pretending an eminently justifiable difference of opinion is just purity signaling is boorish.
Do you think its resonnable that the Gimp project are outputting handfull of human written tests when there is a good chance that they could use a tool to write thousands of tests in an afternoon?
They may be proud of not paying the externalities that the evil megacorps are demanding, but that warm blanket of purity is no protection from being outcompeted.
I think it's a real shame that so many people have set themselves against an emerging technology for reasons that on closer analysis seem like just another round of prejudice and superstition.
Did you read my post at all? I don't think prejudice and superstition enter the picture in the slightest.
From what I've seen, it's less spiritualism and more "they're taking our jobs!" kind of fearmongering. Like all "taking our jobs" fears, whether it's logical or not depends purely on whether they're actually taking these jobs or not.
> You have to serve users, not hold code up as a sacred relic.
They are serving users, by not allowing AI slop. Software developed by people and not an LLM is of much higher quality, and I'm glad they are taking a stand.
The Free Software movement itself is holding up principles as a relic. It has both helped, and hindered the movement in the past. I see no reason for them to stop.
Someone else is welcome to fork GIMP and use AI to serve the users, if they so like.
No one "has to" do anything.
At least DNA has a reward signal to clean up tech dept. Everything is use-it-or-loose-it. Don't use your eyes because it's always dark? Bye bye. Claude would maintain that as a fallback for all eternity.
It wasn't vibe coded, it evolved over millions of years. And we ended up with our eating hole right next to our breathing hole because there's no way up back out of that "decision".
Does free open source software “have to” serve users? You clearly aren’t the user they’re trying to serve. That’s okay though; there’s something for everyone.
> Our own DNA was vibe coded.
No, that was Reinforcement Learning
Neither, if we really want to get technical. RL is directed whereas evolution is not.
It’s frustrating to create tests with models, not sure if it’s everyone’s experience, but every time I asked it to create tests, it makes the tests conform to the sometimes broken code. Rather than calling out what the logic/code is wrong, it will make the tests conform to whatever is already broken.
I recommend iterating on your prompts.
Prod your agent first until it can explain back to you what's wrong with the tests.
Then ask your agent to write a better prompt for writing tests (and save it to an .md file)
Then reference that .md file and iterate on the tests again. It may take a couple of tries, but I've had good results with this strategy.
Ended up with a whole prompting framework.
Prodding an agent vs doing interesting things with my time? What should I choose?
To be fair, some people like to prod. But, for me, I'd rather take hours reading manuals than do trial and error prodding for 30 minutes (even if I knew it would be that short). Yes it's not optimal in terms of time, but it's better for my soul and well being.
> Yes it's not optimal in terms of time, but it's better for my soul and well being.
It is also optimal in terms of time in the long run. The prodder will have to prod over and over again because he has learned nothing. You on the other hand will actually understand what is going on and will be able to either do it yourself or give precise instructions to the machine next time. Don't let yourself become a meat proxy.
It's a good use of your time to ask the agent to reflect and update it's skills if you intervene to manually steer it. Why wouldn't you want it to learn from your guidance?
Perhaps because they're more interested in learning something themselves.
The problem I've had is even when it can explain back the issue, it still writes for itself awful instructions with extra fluff, strange points of focus, and borderline misinterpretations of what I said.
Maybe that's not a big deal, they can still be generally useful, but they seem important to commit with the project if they will be used going forward. I hate the idea of a massive amount of slop riding along with the code, and slop begets slop.
The idea I'm chasing now is writing the documents myself with the hope that enough examples and a document about writing documents will be enough to allow it to write clearly.
It can’t magically know expectations. If the code is well documented or there’s decent associated documentation, I’ve found it has no issue writing tests.
Tests are usually the best documentation and often the only documentation. If humans bother to write anything it should be the tests, not ambiguous human language...beyond the high level 1-liner overviews.
True. It's tough useful to build non-regression tests, that is build tests that test the current behavior, assuming it correct, to possibly spot bugs in the future.
> Rather than calling out what the logic/code is wrong, it will make the tests conform to whatever is already broken.
It is the same like humans: if they say your code is wrong, you will hire someone else who doesn't find so many bugs. It doesn't look good.
Not using AI to at least writes tests in this day and age is crazy work. I started a desktop application 2 months ago and it has almost 3,000 tests now, MANY of them UI tests.
AI is a bit too excellent at writing tests. The number of tests doesn't really convey any information on test coverage. At least I've found that agents that naively improve codebases will write tests for pretty much anything, including irrelevant behavior that has little value in ensuring stability, security or correct behavior. It would probably help to run a test review stage with a strong model, where agents attempt to correct and consolidate tests to guard behavior rather than ensuring that theoretical code paths are followed.
In my experience, designing tests cases to ensure that they're othogonal and verify your invariants (domain or technical) is very mentally taxing and not something you can truly delegate. The actual typing maybe, but testing is a lot of duplicate code and you can go far with just copy-pasting stuff around and building helpers.
Designing for testing helps as well. Pure functions for core logic, coding against interfaces etc.
Using AI to write tests in this day and age is crazy work. The machine is great at writing tests, but the quality of the tests is awful. It will write redundant tests, it will write tests that assert too many thing in one, it will not write tests that need to be tested. It's the part where I have to clean up the most every time. Quantity is not quality, and you really don't want to skip on the quality of your quality assurance. I'm writing a React + Django full-stack application, so there should be more than enough training data out there.
Look, guys, you are literally allowed to fork and go ham on the code.
Yeah if people can make a better and more popular Gimp in an afternoon, go do it.
Gimp has millions of downloads this month. They are doing something right.
[dead]
This raises questions of AI ethics; namely, is it conscionable to subject AI to GIMP’s UI?
> any model released in the last six months would be able to write UI tests covering every single control in an afternoon.
Microsoft also seems to use "models" to "test" its programs. Look how shitty they are.
Funny how GIMP spawned the Gimp Toolkit, aka GTK, which is widely used in Linux, and yet GIMP's own UI seems to have long languished as a rather outdated design. Glad that they have tests, though, and hope they continue to improve it.
Ah! I always thought it was Gnome Toolkit, didn't even bother to check. Thanks
Well it is, regardless of the name (I think "GTK" officially doesn't stand for anything now); to quote https://www.gtk.org/ -
> GTK is a free and open-source project maintained by GNOME and an active community of contributors.
and
> GTK and the GTK logo are trademarks of the GNOME Foundation.
It's GNOME Toolkit now, you're not wrong, but originally was GIMP.
[dead]
They made a blog post to announce that they’d written thirteen unit tests? Sure, there’s value in handmade software and all that but GIMP is not really a particularly excellent piece of software - if anything was to benefit from unleashing the agents it would be GIMP.
They must have set up a reusable harness, and it's the most important part.
It'd be really interesting to look back the comparison between projects with strict non-AI policies (e.g. GIMP) and ArtCraft in 2027.
My personal prediction is that ArtCraft will blow all the old non-AI projects in similar fields out of the water and GIMP will be remembered as a thing of the past. But if I got proven wrong I'd be happy too: at least we will know the value of non-AI policies.
GIMP is far from the best FOSS image editor for many use cases. IMO for many simple editing tasks that people usually need unless they're professional photo editors, even Inkscape, a primarily vector editor, is a better photo editor than GIMP.
Inkscape is what made me believe ArtCraft would be the clear winner, btw. It's very hard to imagine using Inkscape ever again since I tried VectorCraft. (And I'm fully aware that some Path Effects can only be done quickly in Inkscape.)
What is the point of image editors? A fleet of AI artists should be doing it.
I’d go so far as to say that in general, adding tests for legacy software before a refactor or rewrite is the perfect time to put the agent to work.
Did you read the article? These aren’t unit tests. They are UI tests which are pretty hard to do well with desktop apps.
Especially given it isn't as easy as using Cypress.
It sounds like they are nearly end to end user interface (UI) tests, not just unit tests.
we forget, but before AI this would have been at least 6 months of work, at least 50 meetings, and at least 7 separate sprints.
The blog article doesn't say they have written thirteen unit tests.
> *GIMP already has unit tests*. A unit test checks one small part of the code, for example how GIMP reads an XCF file. But our unit tests were not opening a real window. So it coult not tell us if a dialog opens, or if a menu item did its job.
> Note that GIMP *already had tests of GUI code*, usually interacting directly with the core code, and sometimes even simulating input devices’ events (pointer movements, clicks, drags…). But nothing can beat interacting with the real interface. Furthermore such legacy tests directly hijacking core code come with their own problems, such as breaking all too often (and not because they detected a bug), all while not being able to detect whole categories of bugs.
> UI tests check the program as a whole, in the same way as you use it.
> if anything was to benefit from unleashing the agents
AI can write tests now ? For its own software ? Can it evaluate the results ? Correctly ? /s
Writing tests never was the hard part of the job anyway.
Lucky you. Making things testable seemed hard to me quite often and ensuring that the tests covered all the difficult bits without effectively mocking it to death so that one was testing almost nothing.
Yup especially if you don't care if they actually show correctness or not
Timely, as I removed Gimp from my Macbook today after successfully testing PhotoCraft [1] for a week (the AI Photoshop clone written in open-source rust which has had an absured number of updates since you last tried it).
It makes me so HAPPY to remove the dinosaur Gimp, and to run "Adobe" software for free (it's good enough for me, maybe good enough for you). I expect to lose Internet points for saying that.
[1] https://getartcraft.com/apps/photocraft
I have a long history with Gimp. I created Gimpshop... the Gimp maintainers did not appreciate that little stunt, I'll tell ya! :D
I have canceled my Adobe subscription as of last month and have fully moved over to Photocraft. Even if development of Photocraft stopped today, I would be able to do whatever I need to get done in Photoshop without paying Adobe or wrestling with Gimp. It's good enough for me as well. I had been using Adobe Photoshop for more than 30 years now and was bought in enough to spend a significant amount of effort making Gimp act like Photoshop. Bye Photoshop, bye Gimp, bye Gimpshop.
Same with VectorCraft. I spent over a decade looking for an Illustrator alternative. Muscle memory and specific workflows won out. VectorCraft is extremely close.
This and also and Affinity announced Linux version. Hopefully I'll never have to use GIMP again. The UI is soooo bad.
Regular users would use a lot more open source software if more of it was like photocraft and less of it was like GIMP.
Great on GIMP for continuing on and attempting to improve things, you'd think the GIMP team killed people's dogs with some of these comments..
Yeah they always get the flak, even though it's not like people HAVE to use Gimp.
I also hated the UI, coming from Photoshop, but what I found is that when I go back to Photoshop, I'm just as lost as I was before switching to Gimp.
I'm glad that they hadn't lost their courage over the years. Gimp is a solid project, and the FOSS ecosystem would lose a real gem if they stopped.
AI bros can't stand the thought that not everyone agrees with them about LLMs. It's like a religion to them, they feel compelled to get everyone else on board. Never mind that if they are right, it will become apparent to everyone without them having to preach.
It has to be a religion, it can't just be that I prefer the quality of AI-assisted software over projects with no-AI policies.
Besides, GIMP has always been unpopular long before AI.
When a group of people relentlessly insist on hijacking any discussion to refocus the topic onto something unrelated but more interesting to them it stops becoming a preference and starts looking more like bullying at the hands of a religion or cult. For instance:
Hey we added UI tests to GIMP
AI bros: YOU SHOULD HAVE USED AAAAAAAAAIIIIIIIIII
This is more complex than a few unit tests
AI bros: YOU SHOULD HAVE USED AAAAAAAAAIIIIIIIIII
Can we please discuss GIMP?
AI bros: STOP PERSECUTING US
Yeah, sure, discussing software development process and tools in a topic about testing is "religious bullying".
So why do you need to pivot every discussion about "software development process" to AI? Do you believe that the GIMP authors have not yet heard the good word? Do you think they need to be saved from their primitive non-AI practices?
There's a difference between discussion and proselytizing. And, yes, that rises to the level of bullying when AI bros relentlessly drag every discussion back to "WHY U NO AI???". It's absolutely bullying when you make it impossible to discuss non-AI related subjects.
GIMP... didn't have ui tests before?
I'm surprised, tbh. I understand that UI is very much an afterthought in GIMP, but I figured it'd still at least have tests.
GIMP doesn't need UI tests, but a more intuitive UI.
Oh, haven't heard from GIMP in a long time, has there been significant UX improvement and have they carved out a niche like Krita yet?
The tests should simply take a screenshot and send it to an LLM with the question "Is the UI absolutely horrible?". If the answer is yes then the UI is working as expected.
"GIMP now has UI" would be big news
[flagged]