It's funny how fast we went from "Token usage leaderboard! Use AI or you're fired!" to "You're not in trouble, we're just emailing your manager to celebrate your success, we're definitely not making you top of the list for the next culling, we're a family"
> [falconetpt] We now have a tacking system where people were forced to send telemetry from codex/claude into the servers and people are auditing each session
CTO did the same on my team. While a the same time chastising some of the more senior engineers for not always using frontier models.
Engineering leadership in orgs is in a weird place right now.
2027 will be the year of model caps & optimization. Local >> cheap >> premium, depending on the task, your role, and expectations. Tokenomics are coming.
“Prove to me you drove $100k with of value with your usage.”
Sum: For visibility/awareness, not restriction. $300 in a day it emails just you, >$500 in a day it emails you and your manager.
That seems pretty reasonable. Like, "let's make sure we aren't throwing away money on dumb stuff." If it were my org, as long as you can show something with it, you'd be fine. If you're spending that much per day creating throwaway vibe-coded PoCs for fun, or images of a pelican on a bicycle, it might be time to re-orient productivity. Seems pretty sensible/sane to me.
I don't think anyone would think twice about this particular policy in isolation if the same people now worrying about the cost of LLMs didn't spend the last 6 months pushing their developers to use LLMs as much as possible (and in some cases, firing their coworkers claiming they weren't needed anymore because of LLMs).
Sadly many coworkers aren't really needed in the AI age.
I am aware of it happening in CMS projects, content translations are now mostly done by AI, and only require a smaller team to cross-check. Likewise image assets, no longer need to be outsourced to design agencies, stored into the image assets database, instead most of them are now generated.
It's funny how fast we went from "Token usage leaderboard! Use AI or you're fired!" to "You're not in trouble, we're just emailing your manager to celebrate your success, we're definitely not making you top of the list for the next culling, we're a family"
Engineering leadership in orgs is in a weird place right now.
2027 will be the year of model caps & optimization. Local >> cheap >> premium, depending on the task, your role, and expectations. Tokenomics are coming.
“Prove to me you drove $100k with of value with your usage.”
Sum: For visibility/awareness, not restriction. $300 in a day it emails just you, >$500 in a day it emails you and your manager.
That seems pretty reasonable. Like, "let's make sure we aren't throwing away money on dumb stuff." If it were my org, as long as you can show something with it, you'd be fine. If you're spending that much per day creating throwaway vibe-coded PoCs for fun, or images of a pelican on a bicycle, it might be time to re-orient productivity. Seems pretty sensible/sane to me.
Things is, most orgs cannot throw random KPIs to target for AI usage, and then complaint too much is being used.
I certainly am not meeting KPIs, because even if I wanted, there is no way I could outsource all my work to AI, which is kind of what is being pushed.
> That seems pretty reasonable
I don't think anyone would think twice about this particular policy in isolation if the same people now worrying about the cost of LLMs didn't spend the last 6 months pushing their developers to use LLMs as much as possible (and in some cases, firing their coworkers claiming they weren't needed anymore because of LLMs).
Sadly many coworkers aren't really needed in the AI age.
I am aware of it happening in CMS projects, content translations are now mostly done by AI, and only require a smaller team to cross-check. Likewise image assets, no longer need to be outsourced to design agencies, stored into the image assets database, instead most of them are now generated.
"Do as much as possible with AI" and "Make sure you're actually accomplishing something with your AI usage" are obviously compatible directives.
Related:
Corporate America Is Starting to Ration AI as Cost Skyrockets
https://news.ycombinator.com/item?id=48335388