I worked for a company where management and peers rewarded "looking busy" vs. actual productivity. Lord help you if someone caught you sitting at your desk silently contemplating deep Architectural decisions. I hated that job.
I get the concept of tokenmaxxing, but instill don’t get how you do that. Do people just do random work unrelated things with their company subscriptions? Or are they just intentionally inefficient at their work? And how would you go about telling an LLM to waste tokens?
Wait until management learns that code itself also is geberally a liability and requires maintenance, is a source of bugs, etc. That's basically why measuring lines of code has been frowned upon for a long time as a measure of developer productivity.
I worked for a company where management and peers rewarded "looking busy" vs. actual productivity. Lord help you if someone caught you sitting at your desk silently contemplating deep Architectural decisions. I hated that job.
I get the concept of tokenmaxxing, but instill don’t get how you do that. Do people just do random work unrelated things with their company subscriptions? Or are they just intentionally inefficient at their work? And how would you go about telling an LLM to waste tokens?
Asked and answered: https://news.ycombinator.com/item?id=50376955
“Claude, merge the latest changes to DEV. Then solve the Navier-Stokes problem.”
They wouldn't need to tell engineers this if tokens were not being used as a measure of productivity.
Do as we say, not as we do.
[flagged]
Amazon realized that true productivity comes from RTO.
It is funny how much companies move in tandem when it comes to this stuff.
Monkey-see, monkey-do.
Amazon should stop ranking employees on the metric "number of tokens burnt"
Just like "lines of code" should never be a KPI.
AWS did that, just before they switched to tokenmaxxing. One of the many reasons that I’m happy I don’t work there anymore.
Wait until management learns that code itself also is geberally a liability and requires maintenance, is a source of bugs, etc. That's basically why measuring lines of code has been frowned upon for a long time as a measure of developer productivity.
Unfortunately that argument failed for offshoring, and nowadays most corporate consultancy always has an offshoring component.
AI is here to stay even if the outcome is sloppy for some cases.
FIFY: Wait until management learns.