Tootfinder

#silentSunday
Sprague Lake, #RockyMountainNationalPark

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.

I should do some work with Authentik, but I keep getting sidetracked to find out where the default background for the SSO page is coming from
https://github.com/goauthentik/authentik/blob/main/web/src/assets/images/flow_background.jpg

“You opened this page. It already knows the following.”
https://sinceyouarrived.world/taken

taken.
A web page that tells you what your browser gave away the moment you arrived. No login, no form, no permission. Most pages do this. None of them tell you.

Some high-profile publications have migrated from Substack to rivals like Ghost and Beehiiv citing Substack's high fees and increased focus on social features (Emma Roth/The Verge)

Writers are fleeing the Substack Tax
Substack writers are getting tired of the platform’s pricing model and limited customization features, bringing them to rivals like Beehiiv and Ghost.

It’s a funny meme, but it is important to remember that the only time Reagan explicitly threatened nuclear war, he thought the microphone was off.
Trump’s threat was a completely different thing. He broadcast it in a way that assured it would be repeated everywhere and taken seriously. @…

FeloniousPunk (@FeloniousPunk@beige.party)
Attached: 1 image

All the models have improved ‘time to first token’, but Claude seems to have forgotten that theres also the time to second token.
Might as well make a cup of tea in the time it takes to get the rest of the tokens flowing.

NFL Draft rumors: Jets rethinking No. 2 pick, Ohio State could own top five, Bain and Tyson falling?

https://www.cbssports.com/nfl/draft/news/nfl-draft-rumors-20…

How The Daily Wire's relevance to the right may have irrevocably collapsed, as it faces intense competition, falling traffic, layoffs, and ideological battles (Drew Harwell/Washington Post)

Inside Ben Shapiro’s MAGA meltdown
The Daily Wire was once ascendant in right-wing media. Now, the “anti-woke” company faces contentious layoffs, ideological battles and dwindling relevance online.

A new filing from Nexstar in the DirecTV lawsuit says the existing restraining order freezing the Nexstar-Tegna merger could jeopardize assets of both companies (Ted Johnson/Deadline)
https://deadline.com/2026/03/nexstar-tegna-merger-blocked-1236770348/
…

Nexstar Tells Judge Aspects Of Tegna Merger “Cannot Be Reversed”
A judge paused Nexstar's merger with Tegna last week.