Nice example of Goodhart’s law in action.
Tokenistic usage of a tokenised metric
Well that’s what happens when you make it part of your performance review.
Reminds me of a job I had where we got scored based on the number of issues closed. Suddenly, every minor typo or style disagreement became its own issue. We’d close hundreds of issues a day wordsmithing on comments.
Telling people “use more ai” was always a profoundly stupid direction. “Ship more stuff faster, use AI if it helps, get paid the same” would be anti-labor but at least sensible.
If selling widgets the customer probably doesn’t care if you used a screwdriver or a drill. They just want their widget. Mandating drill usage is stupid.
That capex number in the article is a big part of it. $200 billion spent to generate $1-2 billion a month in revenue is wasteful, and people at the top are feeling the pressure to justify that expense.
The people who need the line to keep going up rely on other smaller lines to always keep going up.
Usage score? Retarded executive management.
i hope the people pushing this bullshit are identified and will have their water and energy tokens reduced accordingly in our post-apocalyptic ecototalitarian future.
Tax the shit out of data centers. Let them pay for the infrastructure they use and all this nonsense goes away.
At my job they’ve explicitly called out AI Usage and Aptitude as a core requirement for performance. So you are damn right I am juicing numbers up.
You’ll be the first type of human I’d cut off from water and electricity supply if I were in charge of that sort of thing in the future. Sorry. 😢
In that case, you’re failing to diagnose the problem.
The human causing these problems are not the humans doing actual productive work (+/- AI usage).
The ones you look for are the ones who exercise executive control over the entire organization.
Attacking the working class for using AI as directed is like being angry at two knuckles on someone’s right fist because they punched you in the face.
Totally fair, but in another lens if just nobody used it full stop that would help too.
I’d cut me off too
In a way, this is just a reinvention of line quotas from the past. Neither are good for making something of quality.
Shocking!!
The Great Silicon Valley Token Fire of 2026 continues.
Next steps? The bosses start using the AI tools to summarize the usage of the emoyees to make sure they are using it for the right things. Then employees start subtly prompt injecting their bosses LLM to give them more favorable usage scores. And the sea boils as the arms race to see who can burn more tokens to make investors happy.
Next steps? AI does the bosses job and operates the AI tools.
I get why people do this, I feel the pressure mounting at my job too. At some point we gotta draw a line. We’re gonna collectively manifest AI into being a fact of life, just like they want us to. Purposely using AI for dumb junk isn’t rebellious, its exactly what the AI creators want you to do. They want us all to have one big mass hallucination that this shit is as necessary as they say it is.
Again, purposely using your job’s AI for wasteful purposes is NOT rebellious, but in fact purposely plays into a worse future as planned. If you can afford to quit on principle, and your job is forcing AI usage metrics, thats probably a good time to quit on principle.
Problem is right now the market sucks so you put your head down and keep your job.
Yeah. Market downturn is always when they take the opportunity to make things worse. Resisting the fucking is a big part of what creates a better job market. But I totally understand not everyone is in a position to risk it
TBH, I do this at my job too.
I wouldn’t use it for actual work most of the time, but I’m happy to ask it questions about history and do follow-up research, or give it scenarios for my D&D campaign and ask for its suggestions.
They want to find the best tokenist!
It’s what I’m doing.
