At work we will soon be getting access to agentic AI that parses through our gigantic code base and burns through money for very simple requests.
What would be the best way to maliciously comply with the management’s decision for AI usage to be an everyday thing you are expected to do without getting found out?
My idea is spamming the living shit out of the most expensive models for pointless tasks and then insisting that the result is wrong, and work in the meantime. Nobody can complain if my work is still good, but they will notice that the AI budget is not worth it and stop the hype train.
So, don’t just spam it yourself, automate spamming it; or, even better, let it automate spamming itself–and have it spam coworkers and managers with the results. “Hey let’s make a key/cert for the static code analysis tool and the ci/cd software we already have in place so the LLM can handle the easy issues and hotfix broken unit tests for us.” Then ask the AI to write an app/servlet/interface/whatever that will auto-push new/all issues to it for fixing, and give it an smtp server so it can push out emails. Make sure it’s churning tokens for each email individually. Have it try anyone who has touched the file, so it’s churning tokens for email addresses that have long left the company. Have it churn more tokens to send stuff out to everyone when an email address doesn’t exist. Have it detail it’s “thought” process before, during, and after a fix is attempted & pushed, to churn more tokens, and absorb that text into huge logs that no one will ever read.
Obviously any code generated needs to be reviewed, so have it crank out a full review with a narrative description about how the error was found, why it’s really an error, why it chose this fix, what other fixes it considered.
You should be able to nail a fat load of token spend for every little one-liner that way. Especially once it starts generating its own coding errors that trip the loop, so it just recurses in on itself with no actual human interactive audience to get in the way.
Ooh I just thought of another pointless token boost: make the automated API stuff thank the LLM after every interaction.