But is that before or after fees? Does each fund manager take more or less than the cow's dollar equivalent in pasture?

before fees and milk

Google as just as scammy as the scammers. For many years they manipulated auctions to capture more market share.

As a publisher it really sucks. The DSPs must know these ads are shit. It would be trivial to filter them with a bit of AI.

I've asked my Google partnered SSP to disallow ethically bankrupt ads like gambling, alcohol etc. They say "we've asked the partners to block those categories"... and then it just keeps happening. They're either unable or unwilling to help. I had to take ads down for several days at my own expense due to rogue ads. The ads are now in safe frames which doesn't pay quite as well but I'm okay with it.

There is no explicit IAB category for Gambling and Alcohol in v1 of the IAB Taxonomy. What seems to have been informally agreed is that:

- 'IAB9-7 Card Games' also includes Gambling

- 'IAB8-5 — Cocktails/Beer' and 'IAB8-18 — Wine' means Alcohol.

(Later taxonomy versions have proper values but not everyone supports these)

These blocked categories should be passed to all DSPs by your SSP.

If your SSP cannot block them then get a new SSP.

Don't you use ad quality solution? why?

Yes, Confiant

Name one

This. When it's a person the original quote applies. But big companies have millions of data points on everything and are competent in areas they care about like their own profit. So you can assume it's not ignorance. They don't care if a publisher has to spend loads of time resolving issues rogue ads and malvertising because it doesn't affect them and the end user just blames the publisher.

All the friction goes down to the publishers fighting malvertising.

Suppose a secret like some personally identifying information was recently leaked into the main branch. You are legally required to scrub it. How do you fix it?

Ideally a repo needs to reflect the state of every event that happened. Practically there are edge cases when it's just best to rewrite history.

Research it with an LLM. I've ordered all the stuff for Panasonic and Daikin after a deep planning session. It told me what connectors I need that Panasonic needs a level shifter. Both brands have open source libraries for the internal protocols and wiring. I wanted Zigbee so the ESP32 C6 was the way to go. The Panasonic units I have already have Wifi but most companies suck at this kind of thing. I couldn't get the Panasonic Wifi to pair with my router. Daikin also sell Wifi modules and people don't to like them.

Check out: https://github.com/DomiStyle/esphome-panasonic-ac/blob/maste...

Remember that time when NSA director James Clapper reassured us they're not the baddies.

"Does the NSA collect any type of data at all on millions or hundreds of millions of Americans?"

"No, sir."

Easy. One end in each box.

I've run some evals on my puzzle game https://redactle.net/llm-leaderboard

Deepseek v4.1 flash is able to solve it some of the time. I've found it burns through more reasoning tokens than any other model. Google models like Gemini 3.8 Flash are still dominating and is able to one-shot most evals while being the cheapest.

I'm curious what other unique evals people are running.

It's so bizarre having a low score be GOOD. It's like reverse intuition. Shouldn't it be called `score error` or something along those lines?

Great point. I've changed the naming.

Since it has low activated parameter count but huge total parameter count it needs more tokens to move the relevant information into the context.

Thanks for the help. I ran it on high and it did pretty well and got a lot of one-shots in. The reasoning makes a much bigger difference than some other models.

Does it move the needle on high reasoning?

Yes. I've just run it on high and it did a lot better.

A colleague of mine has a strategy game to compare language models, 4.1 scores pretty high in this:

https://clankerbattle.com/