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.
David Cronenberg, master of body horror movies, was born today (March 15) in 1943 (during World War II!)
I saw his films 'Crash' and 'Naked Lunch' early, around the same time, in 1997 or 1998. Since then, I've seen many more of his- I just counted, I've seen 13 of his 23 films, and I've loved all of them. I'm a huge fan. My favorite was Shivers for a while. I'm not sure what it is now. Gotta see those final 10!
Happy birthday, David. We love …
"Tree planting can combat urban heat, but some neighborhoods are falling behind"
#Trees #Environment
https://
On the first day of the #PTSD intensive, we talked about the shooting. I had felt like I was done with that, that it didn't have anything left for me. But there was something still that filled me with rage... that is still confusing and enraging.
It wasn't actually being shot. I wasn't even the possibility of death. I had been prepared to die. I always knew that was possible. It was something else.
I remember Marc Hokoana's face as he pepper sprayed pacifists, smiling and taunting, joyfully hurting people who he knew were refusing to respond. I remember their flags, the kek flag, literally a Nazi battle flag replaced in 4chan colors with the clover 4chan logo instead of the swastika. How many people have been tortured, have died? How much suffering, that these people not only welcomed but celebrated, joyfully participated in.
The cruelty was the point. It was the plan, the plan he posted to Facebook, the same plan as they have always had, of torturing people until someone responds and then murdering them. Inflicting trauma, responding with overwhelming force, showing how "big and strong" they are because they can always escalate.
Try to stop someone from peppers praying people, they shoot you. Shoot back, like Michael Reinoehl, and they send a death squad for you. But we keep standing up, so they keep escalating to the slightest imagined infraction. Now they just murder you for being in a car, for filming at a protest, for existing.
The bar for what justifies murder or torture will continue to move lower until there is no one left, or until they can no longer escalate.
The feeling of helplessness is still not the biggest thing though. It's the joy with which they inflict this on us. That's it. That's the thing.
CW: gun violence, abuse dynamics
https://hexmhell.writeas.com/the-creature-ptss-5-day-1
Why are we having fewer children?
(Interview with Berkay Ozcan, Professor at LSE)
- Couple formation happens at later age
- Women are choosing "careers" and not just "jobs"
- More people choose not to have kids at all
What else is going on? Short anser: we don't know yet
Even in countries providing a lot of support to parents, fertility rate has still declined
Immigration is no silver bullet. It's part of the solution, not th…
I would love to have seen a 15kg echidna
https://www.abc.net.au/news/2026-04-14/ice-age-giant-echidna-fossil-found-in-east-gippsland/106561526?utm_source=…
Ars claiming the AI article that just made up quotes was a one off. Seems legit.
https://arstechnica.com/staff/2026/02/editors-note-retraction-of-article-containing-fabricated-quotations/
#silentSunday
Sprague Lake, #RockyMountainNationalPark
The entire machinery of online discourse around building and creating has been so thoroughly captured by entrepreneurial "logic"
that we've lost the language to describe what it feels like to simply make a thing that helps someone,
give it away, and move on with your life.
I've been feeling this for a while now, and I suspect a lot of folks who have the itch to build feel it too, even if they haven't articulated it.