Tootfinder

VHS #Flensburg bietet einige Sprachkurse an, leider kein Türkisch.
Aber einen 10w Kurs Gebärdensprache-Basics 🤔
https://www.vhs-flensburg.de/kurse/sprachen

I have a weird feeling, as shit gets more and more real, that stuff like our international trade deals will have to be abandoned out of self-preservation (most were entered into to benefit not NZers, but the wealthiest NZers - a form of implicit TrickleDown). I wonder which political party will have steel enough to point it out.

"sometimes 'asshole' isn't enough," jonathan richman reassessing pablo picasso tonight at baby's all right, a line he's probably used before but feeling pretty universal at the moment. what a miracle to be in a room with richman in 2026. (please don't tell him that i snuck a photo.) #brooklyn

So to follow up on this, I've caught it in action. Models, when quantized a bit, just do a bit more poorly with short contexts. Even going from f32 (as trained) to bf16 (as usually run) to q8 tends to do okay for "normal" context windows. And q4 you start feeling like "this model is a little stupid and gets stuck sometimes” (it is! It's just that it's still mostly careening about in the space of "plausible" most of the time. Not good guesswork, but still in the zone). With long contexts, the probability of parameters collapsing to zero are higher, so the more context the more likelihood you are to see brokenness.
And then at Q2 (2 bits per parameter) or Q1, the model falls apart completely. Parameters collapse to zero easily. You start seeing "all work and no play makes jack a dull boy” sorts of behavior, with intense and unscrutinized repetition, followed by a hard stop when it just stops working.
And quantization is a parameter that a model vendor can turn relatively easily. (they have to regenerate the model from the base with more quantization, but it's a data transformation on the order of running a terabyte through a straightforward and fast process, not like training).
If you have 1000 customers and enough equipment to handle the requests of 700, going from bf16 to q8 is a no-brainer. Suddenly you can handle the load and have a little spare capacity. They get worse results, probably pay the same per token (or they're on a subscription that hides the cost anyway so you are even freer to make trade-offs. There's a reason that subscription products are kinda poorly described.)
It's also possible for them to vary this across a day: use models during quieter periods? Maybe you get an instance running a bf16 quantization. If you use it during a high use period? You get a Q4 model.
Or intelligent routing is possible. No idea if anyone is doing this, but if they monitor what you send a bit, and you generally shoot for an expensive model for simple requests? They could totally substitute a highly quantized version of the model to answer the question.
There are •so many tricks• that can be pulled here. Some of them very reasonable to make, some of them treading into outright misleading or fraudulent, and it's weirdly hard to draw the line between them.

Nitter keeps on giving.
Just saw there this meme being shared. By whom, you ask?
By Sarah Palin.
Yes. That one.

Fleet of ‘Flying Ferries’ Will Provide Zero-Emission, Silent EV Boats for Commuters and Tourists Along Norway’s Coast https://www.goodnewsnetwork.org/fleet-of-flying-ferries-will-provide-zero-emissio…

Fleet of ‘Flying Ferries’ Will Provide Zero-Emission, Silent EV Boats for Commuters and Tourists Along Norway’s Coast
Tourists and commuters along Norway’s coast will soon travel aboard “flying” electric ferries—quietly whizzing above the water from Candela.

Claude is absolutely "king" of AI at the moment, i think however Gemini 3 Flash is good enough for most people and a lot cheaper. I have the feeling ChatGPT will lose this race (just guessing).
#AI #Claude #Anthropic

Schöne Geschichte eines wiedergefundenen Films in der Library of Congress #loc:
https://blogs.loc.gov/loc/2026/02/lost-19th-century-film-by-melies-dis…

If Alito steps down (My feeling is that he deserves to be thrown out) then the Senate D's should refuse to confirm any Cheato nominated successor.
Better to let the SCOTUS seat stay empty rather than have yet another trump sycophant.
https://slate.com/news-and…

It Sure Looks Like This Supreme Court Justice Is Getting Ready to Retire
A flurry of recent hints has prompted new speculation that the court’s most reliably partisan conservative is eyeing the exit.

The new emojis are out - at least the ones Apple includes in the next iOS update:
Distorted Face, Fight Cloud, Hairy Creature, Ballet Dancer, Orca, Landslide, Trombone, Treasure Chest
Another emoji originally was on the list for Unicode 17, then postponed to 18, and in October 25, it was removed from the draft. Probably because a certain company in the consortium took offense to it.
U 1FADD