It’s just Occam’s razor since it specifically references “ Today, the sudden restriction of certain frontier models is deeply regrettable.” in the tweet.
It does send an analytics event when you’re swearing based on a keyword filter (something like is_negative:true), presumably as a signal that the model isn’t performing well this session, but who knows?
Different tax loopholes depending on region etc, but basically like this:
I’m a billionaire earning $100M this year.
I owe $40M as taxes for that. (Too much!)
I find a dumb banana painting by a starving artist.
I buy it from him for $1000.
I wait 6 months.
I go to a museum to get it appraised by “professionals”.
I pay the professional appraiser’s wife $50K as a gift.
The appraiser says the painting is now worth $30M!
Wow that’s awesome, I have such a keen eye for art.
You know what, I’m gonna donate this painting to a museum instead because I’m such a patron of art and culture.
Oh, look at that, I get a tax rebate for the value of my donated painting ($30M)
Now I only have to pay $40M - $30M = $10M in taxes on my $100M income.
There’s more nuance to it in practice, but that’s the gist of it.
-----
Edit: For some reason I can't reply to the comments below so I'm gonna do it here.
> That wouldn't explain the price here, since in your scam the whole idea is to buy cheap and donate dear. not buy for 139M
Now we're getting in the details but it's very suspicious for an appraiser to appraise a work of art from an unknown artist at millions. But it's not that suspicious if they take Van Gogh's Starry Night which was previously appraised at $500M to now be valued at $1B. this way the deca-billionaire still gets to save his taxes while appraiser avoids suspicion.
> As far as I know, that's not how taxes work. You can't get a rebate for the amount of taxes you would have paid, you can get a deduction for the amount of money you made.
There are a lot of loopholes in the complicated tax system for the ultra-wealthy, not for us. This video (still a simple explanation in an animated way) covers a few more of them: https://www.youtube.com/watch?v=dHy07B-UHkE
Interesting they didn’t post any benchmark results - lmarena/artificial analysis etc. I would’ve thought they’d be testing it behind the scenes the same way they did with Gemini 3.
The website does label some relatively harmless elements as ‘dark patterns’, but out of your ‘really bad ones’, I don’t see ‘Competition’ as being a dark pattern.
Competition is a fundamental part of Play. Humans (and other animals) are social creatures and learn via playing and competing with others.
Can people play games by themselves? Yes.
Is competitive play bad or a dark pattern? Not at all.
Counterpoint: Humans visualize stuff in their minds before trying new things or when learning new concepts. An AI system with LLM based language center and a world model to visualize during training and inference would help it overlap more of human intelligence. Also it looks cool.
Edit: After seeing your edited (longer) comment, I have no idea what you’re talking about.
Seems reductive. Some applications require higher context length or fast tokens/s. Consider it a multidimensional Pareto frontier you can optimize for.