spider-mario

Starting at 0 is really only useful if you’re plotting a ratio measure and not just an interval one (https://en.wikipedia.org/wiki/Level_of_measurement#Interval_... ), which I’m not sure this “intelligence index” is.

Nah, it reads much more like ChatGPT than like Claude to me.

> Claude is genuinely good. Better writing than most of the alternatives,

Is that really how people feel? That’s not my subjective experience at all. I find Claude’s writing style “annoying”, for lack of a better description.

Right, but not a model you have already downloaded onto your machine.

Try 3.8 27B in MTPLX; I get about 30 tok/s with the same hardware as you. (Although it does use around 90-95W of power, compared to the ~60W that 3.6 35B-A3B uses to generate 55 tok/s. That’s about 3 J/tok instead of 1.)

Nokia buying Qt in 2008 brought an LGPL option.

It also means you can’t easily know the set of formats that the server could respond with, which in turns makes e.g. archival more difficult.

Most JPEGs are slightly smaller as progressive. The drawback is rather that it’s slightly more resource-intensive to decode, and if you overdo it and split the chroma, you can get weird effects: https://cloudinary.com/blog/progressive_jpegs_and_green_mart...

macOS and Windows’ respective file explorers and built-in image viewers both support it, as do Photoshop, GIMP, Krita and a few more. https://en.wikipedia.org/wiki/JPEG_XL#Official_software_supp...

Two builds of the same source code don’t always produce exactly the same binary either, but we don’t use that as an excuse to distribute object files and call it open-source.

It’s not the “natural conclusion” at all; it’s a fallacy of relative privation.

The point is to parody the pathologisation of cognitive functioning that differs from the societal norm (see: “neuronormativity”). You’re just reasserting that pathologisation.

> "What if I'm the normal one and the rest of society is wrong"

You do realise they’re not saying that in earnest, right? The point is to demonstrate the absurdity of it.

> But eventually you have to reconcile yourself to the reality of autism: it really is a neuro-developmental disorder and you actually do have deficits in social understanding, which you might be able to improve with deliberate effort.

The reality is that what you just wrote is just as true in the non-autistic → autistic direction (https://journals.sagepub.com/doi/10.1177/1362361320919286). Blaming one side for this bidirectional mismatch is not a universal truth.

https://neuroqueer.com/throw-away-the-masters-tools/

“If we start from the assumption that neurotypicals are ‘normal,’ and Autistics are ‘disordered,’ then poor connections between neurotypicals and Autistics inevitably get blamed on some ‘defect’ or ‘deficit’ in Autistics. If an Autistic can’t understand a neurotypical, it’s because Autistics have empathy deficits and impaired communication skills; if a neurotypical can’t understand an Autistic, it’s because Autistics have empathy deficits and poor communication skills. All the frictions and failures of connection between the two groups, and all the difficulties Autistics run into in neurotypical society, all get blamed on Autism. But when our vision is no longer clouded by the illusion of ‘normal,’ we can recognize this double standard for what it is, recognize it as just another manifestation of the sort of privilege and power that dominant majorities so often wield over minorities of any sort.”

> As for autism: try telling someone with severe autism, someone who can't speak or clothe or feed themselves that what they are living with is just another perfectly valid neurotype.

As for womanhood: try telling someone with severe endometriosis or menstrual pain that what they are living with is just another perfectly valid gender.

^ Bulverism.

> Even though the condition is defined in the context of reduced functioning in society.

Diagnostic criteria are not a definition. Even the organisation that publishes the DSM acknowledges it: https://psychiatryonline.org/doi/10.1176/appi.ajp.20250876

> Mental disorders are clearly not “natural kind” categories that can be defined using polythetic symptom checklists, which assume that a symptom can only be “present” or “absent” (9). In fact, it is debatable whether any medical conditions have this property. […] Thus, clinicians use the DSM framework not as the final word on psychiatric nosology but rather as one tool among others for treatment selection and communication with patients and families (11, 12).

In fact, in the case of autism, and if autism is conceptualised as a lifelong neurodevelopmental difference, the criteria logically can’t be the definition since they reference “current functioning”, which could go one way or the other for the same person depending on their circumstances.

What does your point look like when you remove the presumption of shortcomings? I feel that it’s you who might have missed the point of the original post, unless you were just making an unrelated point because neurodivergence happened to be mentioned.

> Neurotypical individuals often assume that their experience of the world is either the only one, or the only correct one. […] NTs are often intolerant of seemingly minor differences in others. When in groups NTs are socially and behaviorally rigid, and frequently insist upon the performance of dysfunctional, destructive, and even impossible rituals as a way of maintaining group identity.

Neurodivergence does not require a diagnosis. https://sherlocksflataffect.tumblr.com/post/121295972384/psa...

And looking outwards is exactly what it does since it references societal neuronormativity. It’s not about the “shortcomings you have”.

Neurodivergent is absolutely not a synonym for autistic.

https://sherlocksflataffect.tumblr.com/post/121295972384/psa...

https://neuroqueer.com/neurodiversity-terms-and-definitions/

Your response seems to skip over exactly the distinction that the parent comment was making. The LLM itself may not have memory or capacity to learn, but the fictional character created by the LLM does. (See also: in-context learning.)

Will this also be applied to older models like Qwen3.6? 35B-A3B still has its uses with its higher speed than the dense 3.8 27B.

> The person on the other side has the same tools you do.

Strictly speaking, not true. I have Gemini Pro; my friend doesn’t.

Where do those estimates of 0.001 Wh/token come from?

> With “a” vs “an”, the pronunciation is more important than the spelling.

I know, hence my question about the pronunciation. If I had thought the spelling was more important, the “X” would have settled it so I wouldn’t have asked.

Or runtime CPU detection and dispatch to support both.

> For instance, we’re quite proud that a XCF file made by a small company for their logo in 1998 still renders the same way in the latest version of GIMP

How do they pronounce “XCF” such that it’s “a XCF file” and not “an XCF file”? “Xeceff”?

RAM is not “never” the issue. My iPhone and MacBook Air could both run larger and more capable models if they had more RAM.

> As capable as it is, it's hard to justify using it when a competing model (e.g. Gemma4:26b-a3b) can consistently achieve the same or similar response with only 1/10th as many 'thinking' tokens, achieve much higher tokens/second, and take a small fraction of the time.

Not my experience at all. On my M1 Max MacBook Pro, Qwen 3.6 35B-A3B provides better responses in similar time (~60 tokens/second, vs. 50 for Gemma 4 26B-A3B). 3.8 27B is slower, being a dense model, but so are dense Gemmas.

“After all, He-Who-Must-Not-Be-Named did great things – terrible, yes, but great.”

> Ads are an attack aimed at your computer too. They download malicious code to spy on you and gather intelligence data about everything you like, you visit, you buy,... it is malicious software hacking into your computer.

The ads themselves don’t do that. Some networks running those ads might.

“Guys, guys, guys, can we take a step back here? What problem are we really trying to solve?”

https://medium.com/conquering-corporate-america/10-tricks-to...

At the same time, though, it doesn’t mean you can’t infer the location parameter of a Cauchy distribution with more and more precision as you receive more samples. It just means that averaging the samples is not the way to do it.