> Find unused code in our web app. Exclude generated files and test fixtures, and check for indirect references before recommending a removal.
If I "/goal remove unused code" in Claude today, I would not even think to specify "check for indirect references" and "don't consider generated code dead code," those sorts of intuitions have been "built in" to the frontier models for a while now.
So I'm pretty confused by this, what does it do that I can't with an agent swarm?
Yep, we've been doing this "agenticly" at my job for a while now. I can't say the results are often impressive, particularly if you underspecify what you actually want.
> Find unused code in our web app. Exclude generated files and test fixtures, and check for indirect references before recommending a removal.
If someone came to me with these requirements, I would laugh in their face. These reek of busy work with no actual business benefit. Have these people actually built real software before?
I would categorize these all as grunt work, but definitely not busywork or even a lack of business value. Improving SEO has direct business value. And the fact that it's unglamorous and laborious feels like a really good reason to have agents do it.
Their sample prompt takes me back a year or two:
> Find unused code in our web app. Exclude generated files and test fixtures, and check for indirect references before recommending a removal.
If I "/goal remove unused code" in Claude today, I would not even think to specify "check for indirect references" and "don't consider generated code dead code," those sorts of intuitions have been "built in" to the frontier models for a while now.
So I'm pretty confused by this, what does it do that I can't with an agent swarm?
Yep, we've been doing this "agenticly" at my job for a while now. I can't say the results are often impressive, particularly if you underspecify what you actually want.
> Improve SEO
> Reduce maintenance overhead
> Make your application compile faster
> Find unused code in our web app. Exclude generated files and test fixtures, and check for indirect references before recommending a removal.
If someone came to me with these requirements, I would laugh in their face. These reek of busy work with no actual business benefit. Have these people actually built real software before?
I would categorize these all as grunt work, but definitely not busywork or even a lack of business value. Improving SEO has direct business value. And the fact that it's unglamorous and laborious feels like a really good reason to have agents do it.
Have you actually built real software before?
In what world does improving SEO, reducing maintenance overhead and faster compile times not provide any actual business benefit?
> If someone came to me with these requirements, I would laugh in their face.
My god the arrogance is almost tangible. Maybe tone down the greybeard act a bit, yeah? No one’s buying it.
just another prompt right?
With a forloop
Yeah, but now with “agentic mapreduce”! A “talent dense” team of prompters selling to the Feds.
which ai come up with that buzzword lol