lambda

Yeah, the problem is, someone has to start the union. It takes work. You have to get the whole workforce to vote on it. A lot of software engineers believe (or believed) that they were too smart and professional to need a union. And of course, if management gets wind of it while you are working on setting up the vote, they may do all kinds of trickery, legal or illegal, to block it.

It is possible. But it's quite uncommon in this industry in the US.

And given the current administration, it is hard to trust you'd get fair enforcement of labor laws if the company did illegal things to block unionization.

I don't know that unions would necessarily negotiate for no use of AI. The SAG-AFRA deals don't preclude all use of AI; they just requre consent and negotiation in certain cases.

A union doesn't give you unilateral power; it just gives you a better seat at the bargaining table. Capital pools its resources to negotiate better as a single bloc; why shouldn't labor as well?

Many screen and voice actors are unionized, and the unions have been striking and bargaining specifcially over these points.

For example: https://sites.suffolk.edu/jhtl/2025/10/30/game-over-for-unau...

Software developers, unfortunately, have been convinced that they don't need to unionize, so have no collective bargaining power for dealing with situations like this.

I'm not just trying to score points about whether he sucks, but specifically point out the types of policies that he advocates for, and there are a number of people in this thread trying to claim somehow that he doesn't advocate for policies that he clearly has been.

So I'm curious, what makes you think that what he calls for isn't ethnic cleansing?

From Wikipedia (and please, fill me in if you have a better definition):

"Ethnic cleansing is the systematic forced removal of ethnic, racial, or religious groups from a given area, with the intent of making the society ethnically homogeneous."

From DHH:

Calling for the deportation (forced removal) of a specific ethnic group (also note the violent analogy to wolves): "When wolves get out of control, you shoot them. When gypsies take over public spaces, you deport them."

Calling for the forced removal of millions of people of various ethnic groups he considers undesirable from Europe: "Because the sacred cow here is not just that mass migration must stop going forward. It's also that millions who are already in Europe must go. Remigration has gone from a fringe concept to the mainstream discourse in record time on account of that realization. The Overton window is swinging wide open, but the cow is still there."

Talking about how much he wanted to move to London but that he no longer has that desire due to its ethnic diversity: "As soon as I was old enough to travel on my own, London was where I wanted to go.... That was then. Now, I wouldn't dream of it. London is no longer the city I was infatuated with in the late '90s and early 2000s. Chiefly because it's no longer full of native Brits. In 2000, more than sixty percent of the city were native Brits. By 2024, that had dropped to about a third. A statistic as evident as day when you walk the streets of London now.... Recently, a projection that Danes would be a minority in their own country by 2096 caused an enormous stir in Denmark. Politicians across the spectrum decried what a catastrophe that would be for this world's oldest continuous monarchy. But a demographic nightmare worse than that has already enveloped London!"

As you can see, DHH is explicitly calling for the force removal (deportation and "remigration") of ethnic groups, because he is afraid of the results of the population becoming less ethnically homogeneous. I don't know how you could argue that what he's calling for is anything other than ethnic cleansing.

Also, note that none of this is due to any actualy on the ground experience; he lives in LA, he's lived in the US for over 20 years. This is purely an exercise in ideology. In fact, LA has a smaller percentage of the population that is white Anglo-American than London does of white British people.

Anyhow: I have laid out in detail the arugment for calling what he calls for "ethnic cleansing". Do you have any argument that it isn't ethnic cleansing that he's calling for? It seems pretty clear to me, so I'm confused why you have kept on insisting "if this is what you think he's arguing for, you either need to re-read the blog post", because I've read all three of these posts of his and more, and it all points towards him advocating for ethnic cleansing.

Tried out the simulator and was surprised to find that the movement keys are ZQSD; then realized that's the equivalent of WASD on an AZERTY keyboard. Checked and sure enough, Pollen Robotics is a French company.

They may want to at least add a preference for keyboard layout, pretty sure that QWERTY and QWERTZ are much more common worldwide than AZERTY.

1. Oracle

2. Microsoft

3. Google

Are probably the worst of the realistic options for acquiring HuggingFace

He uses a word that is a slur for the Romani, an ethnic minority in many European countries. He never discusses nationality.

He also compares deporting them (to where, you might ask) to shooting wolves.

It's also not his only post on the topic. He has previous ones where he complains of how horrible London is because white British people no longer constitute a majority. He calls for not just ends to immigration and closing borders, but "remigration".

It is very clear from these posts that he strongly believes that non white ethnic groups should be removed from European countries.

If that's not ethnic cleansing, I don't know what is.

Nvidia releases some of the most open open weights models, Nemotron 3, which have the full training code open, and most but not all of the training datasets.

Nvidia is a big company. They are good about some things and bad about others.

I think they really do like open weights because they make some of the best hardware for training, and the more open weights models there are, the more people are training and fine-tuning them, mostly on Nvidia hardware.

I feel like Nvidia is one of the better choices for buying Huggingface. Not perfect, but definitely far from the worst.

The word he uses is a pejorative for the Romani, an ethnicity.

Just to explain because not everyone is necessarily a native English speaker or gets the reference.

This is a reference to a dog whistle. Above 25kHz are sounds that dogs can hear but humans generally can't.

The term dog whistle is used to refer to phrases that racists use to indicate their racism, while sounding plausibly deniable. The idea is that it's a phrase that will make other racists realize that you are one of them, while not sounding openly racist to others.

DHH himself has gone beyond dog whistles. He's directly compared what he considers undesirable ehtnic populations to wolves who need to be shot. He's just openly calling for ethnic cleansing by now.

But support for Omarchy itself can be seen as a dog whistle. It's widely known that DHH is a fascist, but you have some plausible deniability by donating to Omarchy because you're just politically neutral and think it's cool tech, or something like that.

He literally calls openly for the ethnic cleansing of undesirable populations from Europe. Just read his blog. There's no subtlety or nuance.

Better than Olmo in performance, not quite as open; release some, but not all, of their datasets.

I specifically use Android because of this, among other requirements, that Apple imposes on software development on their platform.

I do not own general purpose computers that I am not allowed to develop software for without permission. I have always avoided consoles for that reason as well (Steam Machine and other similar platforms would be fine, but I've been avoiding consoles for long enough that it's not something I really look for any more).

I was a major Apple fanboy up until the iPhone. Left the ecosystem after the iPhone and macOS started moving in that direction as well.

I'm going to miss having a smartphone that I can use with my banking and EV apps, but probably for the best to get out of Google's ecosystem. Hoping that GrapheneOS will still allow me to use some of the apps that I like.

Can you provide some references to the research you're referring to? Sounds interesting but you haven't really provided enough information to find it.

They do maintain the Transformers library which is pretty much the core library for how you interact with LLM models in the open source world. So while they weren't using a model they've trained, they were a part of making just about all of the open models (maybe excluding OpenAI and Google's, I wouldn't be surprised if they have their own frameworks that predate the Transformers library).

3.6 Flash scores exactly the same as 3.5 Flash on the Artificial Analysis index. Better on some tasks, worse on others. Mostly within what I'd consider the noise window. Looks pretty much indistinguishable from 3.5 Flash, at least on these benchmarks: https://artificialanalysis.ai/models/gemini-3-6-flash

> I think that in 10-15 years, we are going to have consumer PCs (and phones!) running models doing pretty much anything that frontier models can do right now

10-15 years? The current rate is closer to 10-15 months.

15 months ago, the top model on the Artificial Analysis index was GPT-o3. It scores 30 on the Artificial Analysis index.

Today, you can easily run Qwen 3.6 27B on a variety of consumer hardware. It scores 37 on that index.

Here are a number of open weights models that you can run locally compared with the frontier class models from 7 to 15 months ago: https://artificialanalysis.ai/?models=o3%2Co3-pro%2Cclaude-4...

I've run all of these models on my laptop (Strix Halo, 128 GiB of unified RAM); the bigger ones, like MiniMax M2.7 and DeepSeek V4 Flash, need to be done at fairly aggressive quants that will certainly lose some performance and not quite hit the performance of the unquantized models. But still, it's definitely the case that you can run models that are competitive with the frontier models of 10-15 months ago on consumer laptops.

Heck, just announced though the weights haven't yet been released for independent confirmation is MiniCPM5-2B, a 2 billion parameter (small enough to run on your phone) model, that according to their benchmarks has performance competitive with GPT-4o, a frontier class model from 2024.

https://nitter.net/i/status/2079088670804767114

So that's around 1 year for frontier to consumer device class, 2 years from frontier to phone.

Now, this kind of rate won't necessarily keep up; it's possible that local models will hit a performance ceiling before frontier models do. There's only so much information you can cram into a certain number of bytes, and the AI boom is causing hardware prices to skyrocket so keeping consumer hardware from advancing quite as fast as it had been.

The article describes it.

Both Codex and Claude Code have it, but they work slightly differently.

Claude Code uses Haiku to read through the transcript and decide if the goal has been completed. If not, Haiku injects a prompt back to the main model to indicate what still needs to be done.

In Codex, instead it's a tool available to the main model, plus some part of the surrounding harness that will re-prompt it if the tool calls haven't yet indicated that the goal is complete.

The issue that they are trying to solve is that sometimes models will stop before they have actually fully completed whatever task they were given; attention isn't perfect, and someitmes they'll complete part of it but not the whole task. Rather than making the user come back and re-prompt to keep going, they add a way to automatically do a bit more nudging to try to get the model to finish the task.

I've tried doing a loop of rending the SVG and then tweaking based on that, with local models (so, not nearly as strong). It wasn't very successful; it would mostly report that the image looked great and didn't need any tweaks. Maybe I should try it again, there have been some newer models since I first tried it. And yeah, maybe worth trying with bigger models. But I have found that models aren't necessarily the best at visual reasoning and review, even with a vision loop. Their lack of visual reasoning is part of why they still have trouble with things like ARC-AGI-3.

You can always ask them to draw something else, as a way to avoid any possible pelican related data contamination; given how popular the pelican test is, I'm sure there's some pelican SVG drawing in the training sets of at least some of these models by now. For instance, you could ask for an SVG drawing of a cyborg bear riding a rocket powered unicycle.

It's a silly fun little benchmark, and because Simon's been doing it for so long, you have a lot of examples over the years to compare. But you can always come up with and run your own test with other drawings.

It's really frustrating, because now when I want to write something like a "not X but Y" or "you're absolutely right," I have to stop and decide if I want to self-censor to avoid sounding like a bot.

Sometimes those constructs are actually useful, but man has their overuse really killed them!

It drives us crazy because everyone is using the same 2-3 different machines. So rather than each person having their own unique speaking style, the whole world (or, everyone that publishes direct LLM output) is now speaking in the same couple of styles.

And these machines all tend to converge on very similar styles; they have huge amounts of overlap in training data (much of it being already obnoxious internet marketing), they frequently train on each others outputs, and the RLHF process has a tendency to emphasize certain kinds of "cheap win" styles of speech.

He is saying that weaker models, as measured by a benchmark to distinguish "frontier" models, would be exempted. So an academic lab or startup that isn't yet producing frontier models would be exempted, but once it crossed some benchmark based threshold it would be subject to this kind of oversight.

Of course, right now you've got benchmaxxing going on; some companies specifically targetting benchmarks to appear stronger than they are on a wider range of tasks. Now you might see bench sandbagging, specifically looking weaker on certain benchmarks to avoid regulatory oversight.

For instance, once way I could see this going for open models is to release them undercooked; stop the RLVR process a bit early, leaving them a bit weaker on tool calls and agentic performance, but also release the RLVR environment so people can finish the process themselves.

In fact, this is fairly close to what Nvidia is already doing, the Nemotron 3 models are somewhat undercooked but they are releasing their full training pipeline, to encourage people to use these models as a base for further training, which will generally be done on Nvidia hardware.

Yeah, Kraken SDR removed some functionality due to these concerns, if I remember correctly.

Odd, because export controls don't generally apply to published material (like open source software), but maybe they were worried that because they were also selling the hardware they could have issues due to the combo being export controlled.

Ah, found discussion of what exactly it was they pulled, it was the passive radar code: https://www.reddit.com/r/RTLSDR/comments/yu9rei/krakenrf_pul...

And indeed, they confirmed that they believe the open source software should be OK, but they had concerns because they also sell the compatible hardware: https://nitter.net/rtlsdrblog/status/1591657740229046274

But there are already benchtop or handheld signal analyzer for that purpose.

This seems more like a tool for checking across entire large assemblies like an entire building, car, aircraft, etc, for unknown sources. If you have an individual discrete device that you're already testing, just using traditional instrumentation seems reasonable, but on a large, complex assembly, I can see it being useful. Also useful for things like detecting if a particular antenna is working without actually going up there to measure near it; if you have a MIMO setup with multiple antennas, this might make it easier to check if all of them are working correctly when mounted in inconvenient areas.

I think that for a single device, this probably wouldn't help much over just having a more traditional signal analyzer, either benchtop or handheld. If you know what you're testing, just using a signal analyzer around it will give you a good first pass picture of emissions, and probably be much more informative and precise than this.

This seems more useful for finding unknown or hidden RF sources, for instance looking thorugh an entire building to find unknown RF sources, or maybe a whole complex assembly like a car or aircraft.

This is really weird. Someone creating 4 new accounts just to call this project fraudulent because it can't statically analyze every property you'd like? Does this person have a personal grudge against the author, or something?

Tried it out. I'm compring against Qwen 3.5 122B-A10B, so a much larger model. It gets some correct, but Qwen 3.5 122B-A10B has done much better. Gemma 4 12B even hallucinated some species in trying to identify a plant, and the other guesses it made weren't all that close, while Qwen 3.5 122B-A10B got it right on the first try.

12B did get one right that 31B got wrong. I'd have to do a much more thorough eval to really compare, just a few anecdotal observations and it's kind of hard to really distinguish, but from the samples I've seen, Qwen 3.5 122B-A10B is doing much better at this task.

The 12B architecture definitely is interesting, and it may punch above its weight due to this (though again, would really need to do proper evals to compare). But of the models I've tried, Qwen3.5 122B-A10B really seems like the best for this kind of task.

I haven't run a proper eval, but I've been getting better luck with Qwen models than Gemma on plant and animal identification using vision.

I do like Gemma for translation, however.

How could the harness fix this? It's the jinja template used by the inference engine to render the API requests into the raw text that gets tokenized and completed by the model. Unless you're using something like the raw completions API instead of the `/v1/chat/completions` API, and effectively applying the template yourself. In which case, you could also just modify the jinja template on your server.

Anyhow, I've heard mixed results on any method of supplying reasoning traces beyond the current turn to models not trained on them. For some models, I've heard that it works fine this way, for others I've heard it degrades performance. But I don't know of anyone who has any kind of reliable benchmark for how well this works.