> I wrote this comment the same way. The LLM fixed 14 grammar mistakes that I think would distract readers more than any LLM-ish phrasing.
I don't think that assumption is correct. As you can see by the discussion we're having here, the LLM "fixed" text is actually quite distracting, while text written by a reasonably proficient non-native speaker is generally perfectly readable. It's only if your English is extremely poor to non-existant that it makes more sense to use machine translation or editing rather than writing it yourself.
One problem is that people are becoming quite sensitive to slop, where people just post completely unreviewed, AI generated text. It's quite frustrating, because it's asking readers to read something that no one has ever bothered to write, and it frequently crowds out discussion that people are more interested in. So everyone is kind of hyper-sensitive to signs of AI written text right now, which means when you start to see such signs, your brain moves over to trying to interpret whether it's AI generated rather than reading the text itself.
One thing to keep in mind is that a lot of non-native English speakers use LLMs to translate to English, or to polish their English prose; they may not realize that it causes the translation to come out in a very LLM-style tone. Not sure if that's the case here, but it looks like OP is a native Chinese speaker so may be using tools to translate to English.
But: this was never a problem and now we have to distinguish between LLM generated, human generated, LLM polished and human generated. I'd much prefer it if people just wrote their own text, warts and all.
It's GenAI. It does something that's kind of like what you asked it to do, but it will skip some details or add other ones or whatever.
Dreary architectural pictures will be more likely to have electrical boxes, poor materials, etc, so when it moves the buildings from the latent space for cheery bright architectural renderings to dreary wet November architectural renderings, it will be more likely to add some of those details, because that's what's in its latent space.
Yeah - same things I noticed with people enthusiastically using genAI for old photo coloring. Initially it looks awesome, until you realize it can even alter the human face in such a way, that it no longer looks like that person.
My father was really happy with some old photos colored, until I pointed out he does not look like him. Strangely enough he wasnt bothered...
I have looked into it. This appears to be a Firefox bug when HDR is enabled on wayland and the website is using webgl. Firefox looks to be leaking wl_buffer objects which are causing a VRAM leak in the wayland compositor which then causes performance issues in the AMDGPU TTM buffer object management.
Nice dig. Could you share more about how you narrowed it down in the end? Is it a known issue and you just had to confirm it applies, or did you identify all of this yourself?
`perf` to get go from the "it's stuttering" to "it's spending a very long time in the gpu driver". GDB and printf debugging to get to "the sort in the driver is taking a long time because there are an excessively large amount of TTM buffer objects, not because we are calling it too much". I could have made that leap faster, and I will the next time, but this time that step took me a couple of hours. From there it was a question of who is making those buffer objects, and so it was back to GDB to find nothing in sway/wlroots.
That was where I sort of ran out of good ideas. I have never worked with Wayland before. I figured it's a "protocol" so it must have a way to inspect it, and it does. `WAYLAND_DEBUG=1` allows you to dump the wayland messages, which I then manually inspected to find a discrepancy between allocations and dealloctions. That's a client (aka firefox) bug, so I looked through their issue tracker where I found a somewhat similar bug[1]. I reported my findings there.
Since then I've checked out the firefox code (which I've also never worked with before). Back in GDB and the logs, and I think I know what's going wrong. You can read the bugzilla for that though.
Unexpected side quests are so much more enticing when equipped with the skill set to dig deep – but also, much more time-consuming.. Thanks for sharing! : )
I'm guessing by selling fine-tuning, consulting on hosting, and other services? They also seem to be offering their own inference service with their model, obviously as an open weight model that will be commoditized but I'm sure there are some people who'd prefer to buy from the originating lab. But yeah, when you're offering open weights models, your customers are going to be people who want to self-host, fine tune, etc, so they might be offering services for that.
CBS News installed a new editor-in-chief following the above merge and lawsuit related settlement, and she has pulled segments from 60 Minutes which were critical of the administration: https://www.npr.org/2025/12/22/g-s1-103282/cbs-chief-bari-we... (the segment leaked via a foreign affiliate, and later was broadcast by CBS)
TikTok has been forced to sell to an ally of the current administration, who is now alleged to be censoring information critical of ICE (this last one is as of yet unproven, but the fact is they were forced to sell to someone politically aligned with the president, which doesn't say very good things about freedom of expression): https://www.cosmopolitan.com/politics/a70144099/tiktok-ice-c...
Look into Bazel, a very standard build system used at many large tech companies. It splits fetches from build/test actions and allows blocking network for build/test actions with a single CLI flag. No hassle at all.
The fact that you haven't come across this kind of setup suggests that your hundreds of CI systems are not representative of the industry as a whole.
I agree our sample may not be representative but we try to stay focused on the current and next crop of tpuf customers rather than the software industry as a whole. So far "CI prohibits network access during tests" just hasn't come up as a pain point for any of them, but as I mentioned in another comment [0], we're definitely keeping an open mind about introducing an offline dev experience.
(I am familiar with Bazel, but I'll have to save the war stories for another thread. It's not a build tool we see our particular customers using.)
> although we get the very occasional complaint about wanting a dev environment that works offline
It's only occasional because the people who care about dev environments that work offline are most likely to just skip you and move on.
For actual developer experience, as well as a number of use cases like customers with security and privacy concerns, being able to host locally is essential.
Fair enough if you don't care about those segments of the market, but don't confuse a small number of people asking about it with a small number of people wanting it.
As someone who works for a competitor, they are probably right holding off on that segment for a while. Supporting both cloud and local deployments is somewhere between 20% harder and 300% harder depending on the day.
I'm watching them with excitement. We all learn from each other. There's so much to do.
Yep, we're well aware of the selection bias effects in product feedback. As we grow we're thinking about how to make our product more accessible to small orgs / hobby projects. Introducing a local dev environment may be part of that.
Note that we already have a in-your-own-VPC offering for large orgs with strict security/privacy/regulatory controls.
And he apologized to you and handed the domain over to you.
And now you've decided, 8 years later, to blow up your relationship with a number of other folks in the industry, over a shitpost and some mild criticism?
Please, for the love of god, just drop this whole thing. Dredging this up from years ago has lost you a supplier of a popular product already, and a number of customers. I know a bunch of people who used to want to buy from the cool woman owned hacker manufacturer, and now won't touch it with a 10 foot pole after everything you've done.
And please, for the love of god, stop hiding behind Limor when people are criticizing you. You have repeatedly claimed that people are harassing Limor, when every single piece of criticism I've seen here has been directed at you or the Adafruit social accounts that you post through. And stop using your child as a shield as well.
You are a public figure, and sometimes people are going to disagree with you. They might find your use of GenAI image models to be problematic. They may find your over-hyping of drama happening in the open source hardware community to be a bit much. But you know what? That's OK. People can disagree with you, and make jokes about that disagreement.
But claiming that you are being harassed because people occasionally make a joke at your expense, blowing up relationships with your suppliers, driving away customers because you expose emails and deadnames (or legal names, in cases where people go by pseudonyms online), and doing it all while using your wife and child as a shield is not very professional behavior.
Step back, take a deep breath, pick your battles, own up to your mistakes, apologize for the places where you've gone wrong, and stop using your wife and child as a shield, and maybe you can repair some of this reputational damage. But you really need to get some distance from this.
I eat meat (beef, pork, poultry, and fish) maybe three or four meals a week, and probably about 6 to 8 oz per meal when I eat it. So on a per day basis, yeah, I probably eat about 3-4 ounces of meat per day.
But the source you were quoting was about beef alone. So these are people who eat more beef daily than I eat of any meat.
I don't think that assumption is correct. As you can see by the discussion we're having here, the LLM "fixed" text is actually quite distracting, while text written by a reasonably proficient non-native speaker is generally perfectly readable. It's only if your English is extremely poor to non-existant that it makes more sense to use machine translation or editing rather than writing it yourself.
One problem is that people are becoming quite sensitive to slop, where people just post completely unreviewed, AI generated text. It's quite frustrating, because it's asking readers to read something that no one has ever bothered to write, and it frequently crowds out discussion that people are more interested in. So everyone is kind of hyper-sensitive to signs of AI written text right now, which means when you start to see such signs, your brain moves over to trying to interpret whether it's AI generated rather than reading the text itself.