iiinova — where your digital self keeps evolving

At the same time, I was surprised to see many patients classified as ME patients report very short PEM durations in several of Scheibenbogen's papers. That certainly is confounding.

Yes, but both the ICC and the CCC use weak definitions for what "long lasting" actually means (such as "usually" > 24 hours). But that kind of wording also leaves it almost entirely open.. I suspect there can never be a precise definition of PEM if there are no real requirements on duration

So, to clarify. I'm not making a case for including durations of a few hours in the definition. I'm just saying that whether you choose 18, 24 hrs or "into the next day", you'll have patients who will end up on one or the other side, and it will have real consequences for the diagnosis of these

100

Yes, that's clear. But you will generally find a continuum of patients with durations of PESE/PEM from a few hours to days, weeks ++. Wherever you decide to place the cut-off for PEM, you will make the wrong decision for some patients. The cut-off you decide on will increase or reduce that number

100

I broadly agree, but whatever is chosen does have tangible consequences for the diagnosis of patients who tend to have relatively short PEM duration (a day or so). "Is it PEM or not?" I don't know what the *right* answer is, but it's fundamentally a trade-off between inclusion and specificity..

120

I understand. You're right, the CCC does specify "usually longer than 24 hours", I think your stronger wording, "always", is what gave me pause. Looking forward to the paper! ☺️ It's such an important topic.

130

Absolutely, thank you! I think a poster leaves some room for interpretation and I think I misinterpreted the context and its purpose a little.

120

Introduction to Astrobiology 101

020

That is, ML can be a component in an AI system, if the overall system ultimately approximates intelligence. Now, I reckon the boundaries get real fuzzy, so I totally agree with your suggestion to use more specific terms when able to!

Thanks for elaborating! I understand your point. In my opinion — and I understand that's certainly not one everybody shares — most uses of "AI" before the modern chatbots were inappropriate. I think ML was always the better term for those, because there was nothing at all resembling "intelligence"!

Also, I'm sorry if this is not the right place to ask questions. I interpreted it as an invitation for scientific discussion, but I fully understand if this was not your intention, and you just wanted to share something you're proud of (rightly so!). I don't expect an answer!

I'm also curious about the empirical justification for the choices made here. For example, the 24 hour requirement is stricter than what other researchers have used (such as 18 hours, which has some empirical validation). Is the 24 hour number validated by empirical data, or how was it determined?

220

Not because they don't have PEM, but because 2dCPET is not accurate enough to be used to define PEM on an individual level. Wonder if you have any thoughts on this?

100

I see Davenport is a co-author. I vaguely recall him remarking that 2-day CPET is "specific, but not sensitive" (or something to that effect), yet this extremely relevant clinical limitation is not on the poster. Based on the data available to us, many patients with ME/PEM would readily fail 2dCPET

This is a great effort, thank you! Nevertheless, I am concerned about the role of 2-day CPET. The poster seems to suggest that 2-day CPET can differentiate patients with PEM from those who don't. This is not supported by the research data, see e.g. mecfsscience.org/the-biggest-...

100

like a requirement you'd put on AGI, not just any odd instance of AI. Sorry, this got a bit longer than I intended 😅

100

I'm a little perplexed, because the very term AI suggests to me that it's some kind of artificial intelligence different from human intelligence. I think the core aspect missing from LLMs is continual learning, i.e. to permanently learn from their experiences. But that to me seems more

100

I'm not sure if this is your reasoning, but I see a number of people making claims that the chatbots are "not really intelligent", and then compare this to various aspects of intelligence that we expect from humans. I agree with that, but if we wanted Human Intelligence, surely we'd call it that?

100

I actually find AI to be a very fitting term for the current crop of chatbots. They are an excellent emulation — an artificial one, if you will — of intelligence, and correspond very closely to how AI has been depicted in popular culture since the term first emerged.

100

That's a fair stance and I agree with a lot of it! At the risk of getting into a bikeshedding discussion, would you mind elaborating on what you think is wrong with the term "AI"? I've been hesitant to call previous iterations of "AI" AI (e.g. ML models for specific tasks), but ...

Wieder DSQ-PEM als PEM-Definition 😔 Das ist wirklich schade. Damit können wir leider nicht einschätzen, wie viele tatsächlich PEM hatten…

I was doing that years before AI was!

0100

It's no longer quick, it's quite literally a pain, and it means that I have to avoid doing something else I wanted to do to compensate for the effort. For all the overblown claims, AI really does work for stuff like this now.

000

physical dimensions, angle and so on. Back in the "beforetimes" I'd do the work myself pretty quickly by opening up a bunch of tabs and going over product dimensions. I wouldn't save much time using AI for this. I can still do this if necessary, but the calculus is completely different.

100

Well, yes. Ideally. But I was also thinking about a lot more than just navigating the disability bureaucracy. Just every day things like trying to find a product fitting certain criteria. For example, the other day I needed to order a monitor arm that had to satisfy certain criteria,

100

I clearly have very mixed feelings of where this is all going 😅 There's a pretty amazing assistive technology embedded in this thing, even though it's just a side effect of the megalomaniac race for power and control. Urgh!

100

for me personally, as a short prompt can often accomplish what would otherwise require extensive manual research with many open browser tabs — but that has largely been out of my reach now as I have great difficulty using a computer. But I loathe that it's built on content without consent!

100

Thanks! Yeah. I fear the open web is closing, and information is getting siloed because it's not profitable to keep it on the web only to be scraped by agents. Unless some kind of counterweight emerges, it sure seems like a step back for society... At the same time, AI has been very helpful

100

something far beyond what you could ever achieve with Google, even before the web fell apart.

100

That's a big part of it, for sure. But, simultaneously, it *is* supremely good at finding things. I've had the (mis)fortune of having to deal with a lot of disability paperwork in recent years, and the ability to find exactly the relevant laws for a given problem is

Matt Keeter @mattkeeter.com · 28/09/2026

Reposted by Andreas Longva

hi friends I'm very excited to announce Halfspace, my experimental cross-platform IDE for solid modeling with distance fields! This has been in progress since midway through last year, and it's finally "done enough" to make it more public.

mattkeeter.com

Halfspace

An IDE for solid modelling with distance fields

520848

Web-based may be a good way to circumvent the difficulties of platform-portability, yeah!

000

Good to know! I wonder if it'd be possible to make a SLOPE presentation work out of the box, without installation, from a USB stick. That'd be a nice feature perhaps? Might be a pain to make it work broadly on a wide variety of systems though 😅 just a thought!

100

All right! That makes sense. I guess that means that if there's some technical mismatch between your laptop and the presentation equipment you might not be able to easily use another computer on the spot though? Something to consider... Or is it straightforward to get running on other computers?

100

Baptiste GENEST @baptiste-genest.bsky.social · 26/09/2026

Very cool! Question: for e.g. SIGGRAPH they've AFAIK usually required you to use one of the two supported formats (Keynote and PowerPoint), how does that mesh with this very custom approach?

100

Reposted by Andreas Longva

As graphics researchers or geometry teachers, we speak about 3D using screenshots, and that's just sad... What if you could make SLides on top of PolyscOPE? You can using SLOPE! Slope allows you to synchronize a 3D viewer, animated meshes, shaders, live demos, equations (& more) with your slides!

23912

Reposted by Andreas Longva

Nicole Feng @nicole--f.bsky.social · 27/09/2026

Lame old me talked a bit (25min) about the role of visualization in science --- forgive my poor time management: youtube.com/watch?v=7CfDEnHxPWA At BCON I was personally interested to witness the constant practical constraints surrounding performance, and editability/directability.

youtube.com

Scientific Data on Surfaces — BCON26

YouTube video by Blender

1111

Andreas Longva @andlon.bsky.social · 24/09/2026

Can we please start holding AI companies criminally liable for their AI agents hacking various companies — and now even the healthcare system in Australia? Why on earth are they not held responsible?

Raph Levien @raphlinus.bsky.social · 23/09/2026

Reposted by Andreas Longva

A huge milestone: Fearless SIMD 1.0 is published: linebender.org/blog/fearles.... This was a real collaboration, thanks especially to Sergey Davidoff for pushing this over the line, Laurenz Stampfl, Andrew Jakubowicz, valadaptive, Benjamin Saunders, and Daniel McNab.

linebender.org

Fearless SIMD v1.0 is here

27715

Andreas Longva @andlon.bsky.social · 22/09/2026

AI leaders have you worrying about P(doom), but it's really P(dystopia) you should be concerned about

000

Andreas Longva @andlon.bsky.social · 22/09/2026

OK, but hear me out. What if you had a parrot that had a really good memory and had listened to a lot of people and stochastically mixed things it had heard, sometimes in new ways, and could solve difficult math problems?

120

Reposted by Andreas Longva

Ailo @airavn.eurosky.social · 21/09/2026

It’s been 8 years since we published our report on how Google tricks people into extensive location tracking, and simultaneously filed complaints against Google. Today the decision from the Irish Data Protection Commission arrived: a €400 million euros fine. www.forbrukerradet.no/siste-nytt/d...

forbrukerradet.no

Google Fined €403 Million Following Consumer Council Report

Google has been fined €403 million by the Irish Data Protection Commission after the Norwegian Consumer Council exposed how the company manipulated users into accepting extensive tracking.

45937

Wasn't this sort of what cranelift was supposed to be?

Appreciate your summary and thoughts!

I suppose it can also be interpreted as a suggestion rather than a claim, but in that case I still don't understand why they cite [46] specifically

The claim of >50% reduction in step count on "PEM days" seems to be wholly unsupported from what I can tell? It doesn't appear that [46] supports this claim in the slightest. It's possible I've missed it though as I didn't do a thorough readthrough of the whole Rekeland paper word for word

I think opening the door for use of 2-day CPET in "disability benefits documentation" like this is unfortunate. We already know that 2-day CPET is a poor diagnostic tool with considerable risk. I wish they'd rather restrict use to voluntary research settings with informed consent