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.
Do not coal roll tokens, management will know its you and you’ll probably just hit usage limits without doing anything.
Instead, develop a mission critical program in a custom framework, or even better DSL. The usual, but now vibe coded.
You’re going to learn that from a programmers perspective, ai is far more human than computer.
I don’t know where you’re at, but we’re evaluated on the number of tokens we used as part of our performance metrics. Sometimes you gotta roll those tokens to get those numbers up so you’re not part of the next round of layoffs. Management has no ability to understand token usage as long as the prompts vaguely valid.
I am spared from such nonsense. Meeting bullshit quotas is acceptable but I wouldn’t call it sabotage then.
Introducing code for LLMs to choke on is a great idea. Develop some very specific syntax or structure that training data hasn’t touched, then make it load bearing.
Better yet, find a way to use delimiters, escape sequences, and syntax symbols that are known to interact with Markdown features, so that producing valid syntax for your DSL requires escaping everywhere to avoid fucking up the Markdown LLM output.
Maybe create a module in your repos that are never called by the actual working code, that the AI gives it’s contributions to. It compiles, but never does anything.
This will unfortunately increase the attack surface of your compiled binaries.