Further, DHH isn't calling for the deportation of illegal immigrants in parks. He's calling for the deportation of all people tagged with a specific label with no discrimination. The blog post talks about exactly that difference.

I agree that we live in an ecosystem where, in some industries, all options are terrible. That doesn't mean that, among those bad options, you can't exercise discernment and good judgment. I'm similarly a pragmatist from a different angle - I do believe the marginal dollar value I add to a company's bottom line doesn't change much of anything, and that voting with your wallet only works with collective action to be more than theater - but I also think that makes OP's conscientiousness admirable. Pragmatism isn't apathy or agnosticism, and shouldn't be conflated as such.

If someone insults your ability to do your job, paints you as an opportunistic sellout, and thanks the stars for your riddance, saying someone shouldn't feel hurt because "we can separate personal from professional lives" says more about your inability to perceive your own hurtfulness than it does of that person's benevolence.

The minimum bar for tactful communication is to be courteous and empathetic. None of this had that.

If you meant doing a service job at a small business, where you can have real ownership over how it treats its customers, I would agree with you.

Further, servers still have hosting value, but any business running agents is almost certainly going to want a sandbox that limits what code runs for agentic work, so targeting _sandbox_ environments is probably the better bet long-term. And, yes, you could implement your proposal in any chroot jail or gvisor, but nobody wants to get their hands dirty finnicking with that - programmatic access control beats file-based access control for the simple reason it's managed for you.

If anything, my critique of OP's implementation is actually the opposite of yours: they've chosen the right primitive and layer, but people really need contextual access control rather than RBAC. Sort of like ongoing zero trust. If it was possible to inspect the context, decide if it was a bad idea to allow the tool call, without exposing the decider to untrusted context, you could have something that really changes things.

Is the difference more about measuring a request "across services"? That is, the total cumulative p99 across services must be small i.e. linking all requests to a user and then measuring that? Or is the difference elsewhere?

If the former: are you taking traces and graphing that? What's your methodology?