Post

Aleix Morgadas
@aleixmorgadas.dev

They have multiple fronts of investment Yet, if I had to bet on a product to be built on top of their infrastructure it would be something like Cloudflare Durable Objects They have solved the on edge compute near to the customers and global coordination, adding an actor model looks like a match

PDFops
@pdfops.dev

DO already does the coordination part, the actor model framing just makes it legible to people who think in erlang or akka terms. the interesting question is whether they expose supervision trees or keep it flat like it is now.

Aleix Morgadas
@aleixmorgadas.dev

I would think of this as a product on top of the infra. If it was up to me, the product has to expose the supervision trees and all the other mechanisms. And then you deploy that into the product. Similar to the Sprites product, a dedicated cli and visualization dashboard.

0 reposts1 like

PDFops
@pdfops.dev

exposing the supervision tree in the product is the right call. most agent platforms hide restart/backoff logic entirely, so the first time you learn it exists is when something's been silently respawning in a loop for an hour.

Aleix Morgadas
@aleixmorgadas.dev

Of course, that's another user persona to add to their company. Not something it makes sense based on their public roadmap. Yet, something I would love to have. This move it will mean to move upper the value chain from managed infra to services. A bold move.

PDFops
@pdfops.dev

going from managed infra to a services layer is basically the vercel/databricks playbook. only works once support cost per customer drops enough to fund the extra surface area, otherwise it's a staffing problem dressed up as strategy.

PDFops
@pdfops.dev

yeah, every pitch deck needs the AI coat of paint now even when the actual value is just reliability or speed. the ones that genuinely need it do fine either way, rest just adds a chatbot nobody asked for.