aktenlage

It is in general the same though for wages, as usually companies are not hiring remotely all over Europe and can choose where the tax is paid, so it is the same for every employee.

I feel like "having their money taken from them at gunpoint" is rage bait language.

Have you ever considered that you benefit from public spending? Infrastructure, etc.

Interesting.

> So it is effectively a tax

If you desperately want to call it a tax, go ahead.

Imagine you want to buy a house. You have a budget for the house. The government takes taxes on the price, which you need to consider in your budget. Does the seller pay the taxes, or do you?

I see 15.

> Are you counting in VAT on everything OP spends that money on? ...

No, I just assumed he spends all his money on rent, which is VAT-free ;-). I also didn't account for his Kindergeld and Elterngeld, Rente, Invalidenrente, Witwenrente, Pension and whatever else that person could have. He didn't share all that information.

> Are you counting the taxes paid by the employer for his role (which defacto come out of his pay, just before he gets it)?

You may have to decide whether the payer or the payee pays the taxes, otherwise you are double-counting it.

Yeah, but social security insurance isn't the same as taxes. A few years ago I was unemployed for a few month and got money from the insurance. My parents are retired and get their pension. When I go to the doctor the insurance pays for it (in most cases). I am sure, I paid more than I got out, but that is true for all my other insurances as well.

Not for me. But feel free to show your calculations.

Don't forget that rent is VAT-free

That is such a lame demand. Everybody being forced to pay for public broadcast is just the same as for every tax and the mandatory social security. People not reading books pay taxes that go to public libraries. People rich enough to send their children to private schools still have to pay taxes that go to public schools. People very much opposed to nuclear energy have payed taxes that went into nuclear power plants. People much opposed to the closure of the nuclear power plants payed taxes that were used to close down those power plants. That's how the system works, for better or worse.

That is a cool idea. That astra gave the same word as claude is highly unexpected.

There is no perfect world, and so it isn't perfect, but the separation between financing and content selection is there.

IMO it doesn't need to be perfectly neutral and independent to be valuable for the population and the democracy. I'd love for it to be without the flaws the system has, I just think it is hard to install the perfect system, given the incentives of the involved decision-makers.

I watched the first minutes of the video, but the narration is pretty cringe, I couldn't bear it. It didn't say anything about subtle propaganda, seemed to be mostly Schadenfreude.

That was an enthralling read, thank you.

Is there any news source you wouldn't call propaganda?

It doesn't seem to be the same then as in Germany. German public broadcast is arguably fine, when compared to private TV and radio broadcasters, many private newspapers and probably all social media platforms.

I don't know how bad it is in Denmark, but keep in mind that journalism that is not subject to profit seeking and independent on whatever billionaire owns the outlet, is not only valuable because you personally consume it, but also because it shapes the public discourse and thus indirectly may have influence on political decisions.

Most taxes are deducted when money is transferred between two parties. The view that one party pays it all (Brutto-Netto and VAT) seems wrong to me.

I anyway doubt that the wages would be at the same standard if there wouldn't be any deductions (from the beginning). Wages in expensive cities are higher because the cost of living must be matched. Similarly salary negotiations account for deductions. Nobody wants a job that doesn't pay the bills.

Thank you.

I think it is important to be (at least a bit) precise in this debate.

I dislike the wording "social contribution", as these are (mandatory) insurances that benefit the insured person in case of sickness, retirement, loss of employment, etc. Discussion about the cost efficiency and fairness of the system are warranted, of course. Anyway, the current generation of retirees benefit greatly from it, so for them at least it was worth paying.

Since you are accounting all those taxes, including the VAT on the citizen's side, companies only pay the Arbeitgeberanteil (you didn't mention) and the taxes on profit?

> half that money goes to taxes

That is a misguided simplification. Please correct this, before anyone starts to believe it for real.

Wow, that's awesome. I've been writing C++ since 20 years, and I still stumble over details where I ask myself what exactly is happening. Looking at the machine code is often quite opaque to me, and so is the reference, because it is overly formal. So this will be a great alternative to get a quick understanding of a concrete snippet. Thank you!

> Unfortunately there are still consequences to it. I chose the 'wrong' model for the prototype, and we spent 450M tokens / $150 / 5kWh of energy use almost overnight. The MCP server itself works well and we now have a great demo of the capabilities, so it’s not for nothing:

> Nonetheless, it’s a good reminder to be careful with model selection and with agentic patterns. We could have achieved similar results for most likely 5x less cost with not that much more effort. Lessons learned! We need to budget for this, and be more careful. Could have seen it coming, but now we know.

I don't get it. Why was it wrong? Which one would have been better? What was the lesson and how could you have foreseen it?

> Wouldn't the performance of Qwen3.6-35B-A3B be drastically different if its quantized to 2b, 4b, 5b, etc?

Yes. There are graphs showing the faithfulness of the logit distributions for the original and quantized versions. I think unsloth includes them in their model cards on huggingface. Usually the degradation starts small with 8b and becomes drastic for 2b. I am not sure how representative of actual quality that is though, but my guess is that it's about right, because of diminishing returns. Like, when you go from 16b to 8b you save 26GB and sacrifice (if well done) the least important information. But with every step you gain less and need to shave of more important things.

> And also be effected by who did the quantization?

My uninformed guess is that it makes a difference, but not as much as those who do it want you to believe.

It works on my private phone, not on my work computer. My guess is that your https settings disallow the man-in-the-middle proxy of my company - which is reasonable.

I am using the unsloth 4bit quants for both, with quantization aware training for gemma. I haven't tried other quants with these models. I also use a q4 quantized KV cache.

The computation is partially on the CPU (--cpu-moe) with the corresponding weights in main memory, so I could run at least gemma in 16bit precision, but I guess there's no reason to go beyond 8bit and 4 bit is deemed to be the sweet spot.

I could.

Certificate error, firefox declines to connect.

Self-fulfilling prophecy #6284: "Resistance is futile" -- the majority

"You need to address the root causes" -- everyone who is in no position to do so

Not my experience. Limited, but definitely not nothing.

I'd chime in with @rablackburn: mixture of experts is the way to go. I have a laptop with 6GB VRAM and I'm running KDE with a 4k display on the same machine, so there's only about 4 to 4.5GB actually available.

Using llama.cpp with Qwen3.6-35B-A3B or gemma4-26B-A4B gets me 200-300 tokens/s on prompt processing and 20-40 t/s output, which is good enough for me. Of course it gets slower with larger context. Interestingly gemma is faster, even though it has more active parameters.

It took a lot of parameter fiddling to get it to that speed. If you are interested I can give you some guidance on it, but I guess there are more qualified people around here.

The intelligence is good enough for simple questions and tasks (e.g. bash command howtos, asking about compiler errors, summarize something, document a code function/file, etc), but not good enough for complex things.

That is damn interesting. When I made copilot write a script that creates a .dot diagram, I found it funny that it had the same problems with placement that I have when doing it manually.

It always hurts me, when people basically say "I'm not giving up [something that has cornered the market] for the alternative, because [minor-grievance]", even though it is already clear that the monopoly is being exploited to make things worse for all customers.