Discussion
For decades managers searched for a way to measure developer productivity. Now AI has given them tokens, and they're convinced they've discovered the holy grail.
We spent 50 years trying to get rid of managers counting lines of code just for them to reinvent it with extra steps and a cloud bill.
@fribbledom you didn't really think it would be that easy??
@fribbledom The most interesting question is though perhaps: Why do they need to measure developer productivity?
I mean, really: Why?
@fribbledom why do they always measure "output". Would it be nice when they'd measure "Outcome" instead...
@virbonus
The fact that it has been created by genAI totally undo its point.
@fribbledom
@fribbledom "We have long known that turning a measurement into a metric makes it useless as a measurement, but if we *start* with a useless measurement ... "
@fribbledom In my last job I used to really piss off the managers who talked about token-maxing by referring to it as loot crates.
Install Fedizen
Get the full app experience
@fribbledom
Want a good "way to measure developer productivity"?
How about good old-fashioned "return on investment" (ROI)?
Let's say that someone develops a thing with AI in 2 weeks instead of 12. You've just saved 10 weeks and 10 weeks of their pay for just the cost of a few tokens. But let's say the output is sloppy, performs poorly, and costs 10x the memory to store the AI-generated code. If the product will be produced in low-enough quantity, you might still come out ahead. If your only competitors also don't use human-optimized code, their products could be just as bad, and you'd still do alright. But if someone is able to prove that their code (or other intellectual property) wound up in your product without their permission (through injestion into non-discerning AI) and sues for infringement, then your ROI starts to go way down, potentially going quite negative.
#ArtificialIntelligence