This is really awesome. I tried my test-task (generate a Python wrapper for a fairly complex C-interface) with gemma4:26b-a4b-it-qat. And for the first time it would just do it, without prodding and without errors. (Of course Claude just did it a heartbeat, too.)

My optimal local setup now is gemmma4-qat and Q8_0 K/V cache quantization with 256k context windows. And that runs fine with 12GB VRAM and another 10GB in RAM.

Previously I tried with gemma4:26b-a4b-it-q4_K_M and qwen3.6:35b-a3b-q4_K_M, and they both would tie themselves into knots (especially qwen3.6 can take forever with incessant "but wait..." thinking loops.) More often than not, they would not finish the task.

It seems true these 4b QAT models are as precise as Q8_0 quantization (which is supposedly indistinguishable from bf16).

I am really excited about the prospect of local LLM inference.

Thank god some sanity prevails.

Adding these to the index immediately would force passive index funds (multiple trillions of $) to buy this stock, and thus not allow the market to make performance based decisions.

It's truly a shame that the NASDAQ caved and I will definitely reduce my position in such index funds (I have less trust in it now).

I use Claude every day. Often for multiple hours a day. Basically doing my job not worrying how many tokens I spend (as in too many or too few). This is a pretty complex code base (database optimizer and related).

Just looked at spent for the past 30 day, didn't even come to $600. 95% of my tokens are from cache. If I were to reach even $1500 I have to let claude run unsupervised over night (and with the amount of mistakes it still makes and guidance it needs, I do not believe we are there yet.)

> didn't even come to $600.

That's still in the ballpark. A modest change in your usage habits or workload could easily get you there.

is this with a subscription or pure API billing?

I am using Opus 4.x at work, and these "smaller" (20-80bn, 3-4bn active) models at home. Unfortunately there is no comparison, yet (IMHO anyway).

With Opus I can work, trust its designs, architecture suggestions, and code changes, even in a complex code base.

The smaller models seem to "try". They work for smaller tasks, but for more complex task it's often more work than doing it myself.

I wish it were different, and maybe in a year or two it will be.

Ya when it’s your own token budget on the line the smaller/cheaper models are more attractive.

I’ve used GPT mini quite a bit and it’s decent.

Please, not the SAT!

My son is prepping for the SAT and I am helping him. I studied physics and computer science, and was a advanced math A+ student...

IMHO: The SAT is useless, solving equations under artificial time constraints is something that only happens in these kind of tests. The focus is on solving problems fast and getting a good score, and nobody really cares if you understand the math behind it.

So, please, if you go back to testing, find something more useful than the SAT.

Good.

Just naming things differently does not work in other countries.

If it quacks like a duck, swims like a duck, and looks like a duck, then it probably is a duck.

This is a big part of the reason we have judges, and why we should not replace them with LLMs.

Also the reason laws aren't overly specific, so these judges can make those judgement calls in the cases. Following the spirit of the law rather than the letter of the law is something many judges care about as well.

Counterargument: Any "specific" wording of gambling would have condemned this activity. On average, vagueness in law generates more loopholes for corporate criminals to hide behind than it generates discretion for judges and juries to rule against them.

What is your stance on loot boxes in video games then? Is that akin to a poker/slot machine or is it more like a kinder surprise?

I wouldn't personally lose sleep on whichever way loot boxes were legislated. If their contents are easily converted into real-world cash, I would say the company bears some responsibility for the actions of any black markets, perhaps proportional to the volume traded. CS:GO has to crack down on its massive skins market, Kinder doesn't have to sweat its little toys.

Generally, I think the common-law notion of "consideration + chance + prize" augmented with the requirement of an "insurable interest" (in order to keep the insurance industry legal) discerns the truth well enough. Insisting that any "prize" not be monetary or monetarily-convertible keeps it from becoming a cyclical addiction, and evaluating infractions relative to volume keeps extremists from banning gumball machines and Kinder eggs.

You have an US-centric (an really naive) view of the judicial systems of the world. In many countries the judges have to follow the exact letter of the law. In others, this "spirit of the law" excuse makes way for enabling systemic corruption.

> You have an US-centric (an really naive) view of the judicial systems of the world.

Right, so because "judges have to follow the exact letter of the law" in the countries you have in mind, somehow my view is US-centric, although I'm from Sweden and I live in Spain, never visited US nor been in any US court ever? Both countries I'm familiar with, have purposive interpretation rules.

You can share that in other places it might be different, compared to how it works in the places I'm familiar with, without chucking insults around, especially insults you uninspiringly basically said in every recent comment of yours, at least make it unique.

To actually make this a somewhat interesting conversation instead of just pooping words, as I just tried that and didn't find it as fun as you did; what specific countries are you talking about, how does it work there, and how little corruption do those countries have compared to say Sweden?

You have ni idea what you're talking about if you think there's such a thing as "the exact letter of the law".

funnily enough - "letter of the law" interpretations can be so general that many left US leaning people believe it is abused by the right for systemic corruption: https://youtu.be/l7To2evwGKs?si=i93YDCuqCl5PMPlY

Other countries paying $10,000's to educate people who then want to apply this knowledge in the US. US reaction: "Nah." Besides, we are talking about legal immigration here.

I don't get it.

Are we purposefully dumbing down the country now?

Sure, some things are trade-secrets or national security issues or whatnot, but those are already not shared.

More than militarily the US has always led with soft-power (science, culture, etc). We are throwing all this soft-power away.

Perhaps there is a legitimate reason, but like so many things in this administration this feels like a knee-jerk reaction to... Something.

> Are we purposefully dumbing down the country now?

Yes.

> Wrong the way it would be wrong to predict that if you set your kitchen on fire, the result will be a renovation.

This might be favorite metaphor ever, and one I'll quoting in the future! :)

I think the author conflates social media with other inventions like a portable GPS device, an electronic map, a music player, or indeed a cell phone.

As far as social media goes the author is (IMHO) spot on. You do not have to look far to see how that is at least harming democracy around the globe. For democracy to flourish you need reflective voters who can entertain multiple viewpoints and make informed decisions. That is what social media - in its most common current form - discourages and rather optimizes for attention-time (which is money).

And of course (some) anonymity paired with global reach would not bring out the best in people. Anger and flames spread faster than conciliatory messages and get you more dopamine posting those.

Just my $0.02.

I stumbled over that metaphor. Isn't it true that a consequence of setting your kitchen on fire will be a new, better kitchen?

Well, as a secondary consequence maybe, but then you could not set your kitchen on fire and still renovate it. Supposedly the first step you think of when renovating your kitchen isn't "Let me set my house on fire!"?

Two ways to think of this:

1) Sometimes a incident is the best way to get a project done. Working in FAANG I've seen a project get done in 1 day during an outage that was projected to take MONTHS during normal business.

2) Sometimes that renovation would never happen due to reasons. Sometimes you need some kindle to start the fire [pun intended].

As long as you can convince your insurance company that you didn't do it on purpose to get a renovation, I guess. :-)

Isn't the old adage that democracy was never good, it was always just better than all the other forms of government. It got more done. It advanced economies more. Etc etc etc.

Then we torched it at just about the same time as the Chinese came along with a new form of government that I'm not sure the world has as yet even given a proper name. (I guess we can call it Communism? But everyone kind of knows that it's nothing like.)

So to global generations that have grown up viewing all these changes, democracy by comparison to what they have in China has started to look not so all powerful. To many of the planet's young people the assertion that "democracy is the worst except for all the others", is by no means obvious. That change in view is going to have profound implications on the world going forward.

>Then we torched it at just about the same time as the Chinese came along with a new form of government that I'm not sure the world has as yet even given a proper name. (I guess we can call it Communism? But everyone kind of knows that it's nothing like.)

I think the term is "state capitalism." (https://en.wikipedia.org/wiki/State_capitalism)

This term was how 'true leftists' would separate themselves from the Soviet Union and 'old communism'. So imo, China is something else. (Or as Chomsky says, USA is also state capitalist, so could be anyone!)

>This term was how 'true leftists' would separate themselves from the Soviet Union and 'old communism'.

Soviet communism is different from Maoist communism, which is different from Juche. Every political model has variations in terms of ideology and execution, and they do evolve over time. It is correct to differentiate between Chinese communism and Soviet communism just as it's correct to distinguish between European and US capitalism.

>Or as Chomsky says, USA is also state capitalist, so could be anyone!

I think an argument can be made that the US is headed in China's direction in that regard, yes.

Rigid political taxonomies tend to lead to thought-terminating cliches, which is why they get deployed in propaganda. Reality tends to be more subtle. Socialism can exist within capitalism, and capitalism within socialism. Communism can be authoritarian, and it can be so egalitarian that it collapses (as happened with many communes in the 1960s.) Communists can be ideological enemies in the same way as Christians, Muslims and Jews, despite ostensibly having the same origin. And plenty of self-described free market capitalists would love for America to have free economic zones like Shenzhen.

Whatever China is, it does seem to be more capitalist than communist to me.

China’s ‘new’ form of government is basically their old form of government with some communist rhetoric sprinkled over it.

I know it's (genuinely) hard to believe in this day and age, but pre-2000, and especially before the founding of Fox News in 1996, the news you got was much, much more likely to be genuine, with real investigative reporting that wasn't heavily interdicted by powerful and moneyed interests.

This is not to say there was no bias, nor that the powerful and moneyed interests had no influence—but it was much less than it is today. There was much more of a social norm of news being honest, factual, and relevant.

Who do you think decides which media the algorithms show you now? It's all corporate, just more addictive and less accountable now.

I guess it all depends on what you use it for.

I work on database optimizers and other database related stuff, and I can assure Claude Code - with all the highest settings - does make mistakes. It will generate a test that does not actually test what it "thinks" it tests. It will confidently break stuff.

Do not get me wrong. It is still awesome! It takes much of grunt work off me. It can game out designs decisions even when that needs to refactor a lot of code. If you point out a mistake more often than not it can fix it itself.

It's just for a critical project I would never ship it without understanding every line of code - with the exception perhaps of some of the test code. Maybe in a year or two that will be different.