Nobody cares whether the benchmark is no longer accurate (or whether it was accurate in the first place). People are furious because their investment allocation is tied to this rule change. If you or anyone wants an orignalist SnP number that accurately (by whatever standard you want it to be) benchmarks the market, you are free to do so very easily.
I do that for personal use too (although $2.4k/yr for me because I only have an Claude Max subscription). Outside of my hobby projects Opus also manages my personal accounting, researches and organizes info (travel plan, what to buy and where to buy, etc), helps me reply to emails when I'm working in the kitchen, etc. I consider it well worth the price. Tbh I'm willing to pay more than what I currently do, but competition is good for the consumers.
To put it very simplistically, from mostly a practical view:
abs: cannot be differentiated around 0; has multiple minima; the error space has sharp ridges
power 4: way too sensitive to noise
power 3: var(x+y) != var(x) + var(y)
I use Claude Code at work via AWS bedrock, also personally subscribe to the $20/month Claude. Anecdotallt, Sonnet hasn't slowed down at all. ChatGPT 5 through enterprise plan, on the other hand, has noticeably slowed down or sometimes just not return anything.
Sure, they work in that huge codebase, but for any specific task, the number of lines your team keeps in their short/long term memory is surely much much smaller than that.
Chinese internet mostly consists of a few closed gardens tightly controlled by big corps. Crawlers simply don't work when each company employs an army of engineers to guard their data. Many of the most popular websites are also app only. It's impossible to get the corpus necessary to train a good LLM.
I had to file my tax by postal mail in the US. Granted there is the option to file online, but that only works for ~80% of the people when things are completely within the intended domain. I have just one extra item outside of standard salary slips and some investment income, so I had to file physically.
Why is this correct? Youtube is a private company. It's very much allowed to charge $200 per month while playing tons of ads for you. Youtube sets the price and its policy, not the users.
This is like saying you have the rights to mute your work meetings. Sure you do, but you just won't be employed anymore. I don't see that as a problem, because being employed and watching YouTube are not essential services nor human rights.
I use it because my company bought the enterprise plan and trained a bunch of "specialist GPTs" that were fed internal documents, code bases, Slack threads etc. I am not aware of other good LLM companies providing the same level of integration.
So is Wikipedia. Otherwise Chinese people just cannot stop reading all those wiki pages about that fungi that only grow on a certain volcano in French New Guinea. How addictive!
I don't disagree there's a lot of complications in the actual implementation, but this approach is a better-than-status-quo way to achieve some fairness IMO. Currently the US Dept of Labor has a system of determining the fair wage that should be paid to a certain job description at a certain geographic location. In the green card process this is called "prevailing wage determination". Why not use it for H1b too?