Is there evidence that frontier models at anthropic, openai or google or whatnot are not using comparable optimizations to draw down their coats and that their markup is just higher because they can?

The auth between your app and the proxy can be scoped more easily.

For example if the proxy runs in localhost you can trust the localhost workload.

Or you can use some other kind of workload identity proof (like cloud based metadata servers). If you leak such a key no other VM can use it, because it's scoped to your VM

That's not true. Both AWS' as well as GCP's workload identity tokens are not bound to the VM. If you leak the credentials they're valid until they expire. on AWS the expiry is 6 hours (non-configurable). Even if your IAM role has a shorter expiration, the credentials assumed by the VM will always be valid for 6 hours.

That entirely depends on the location of the proxy and the extra conditions you can express. E.g. you could bind it to a source IP and have the proxy check that, or use some overlay network (like tailscale does)

My point was that you don't literally have to run the proxy on localhost in order to scope the request.

in exchange for service that presumably a) costs something to amazon to operate (so not pure 100B profit) and b) anthropic would have to spend anyway to operate their business.

so basically ...

you could view this as a kind of discount, but instead of paying less later, you get some cash now and then pay full later.

Name checks out

If such tax would introduce an asymmetry that will favour human employment, then there should be enough buying power to create some demand

I wonder what percentage of the job space truly depends on the current edge we have over machines.

I think it's reasonable to worry that way before machines are more reliable than the average human (let alone more reliable than a highly trained human) they can pose a significant disruption to the job market which will send shockwaves throughout society

That is why we need functioning states -- free markets won't save you in such a case. Though I found it is hard to explain especially to U.S. people, who put "regulation" on par with f words :)

Consequential enough that it's reasonable to plan for it.

Just train machines on the huge corpus of human needs so they can need things like no human has needed things before.

What can possibly go wrong

Perfect place for the fortress of solitude!

Not if it's on a map anymore

Yeah. Basically conquest is possible when the victim is weakened. There are many ways to become weakened. Infighting and disease are common causes of weakening.