Agents with MacOS VMs are very powerful. Just one example: Devin can test android and iOS apps side-by-side.

I’m so confused. Why would I pick Devin over Codex that does exactly this as well?

awesome, now hopefully you'll consider to have in between $20 and $200 plan (eg. $100?)

oh snap

What ive done with memory is the future memory is stored as events in the durable shared log. Each memory object has a memory_id. A memory.write introduces that ID, and subsequent lifecycle events reference the same ID. On recovery the MemoryStore verifies the log, clears its in-memory projection, reads the events in sequence, and rebuilds the dictionary of memory objects keyed by memory_id. The reconstructed status comes from replaying those events—not from trusting whatever current state an agent claims.

crazy, already taking care of last mile problems

Y'all are tempting me...

this is next level and this is the future. You guys nailed it with this one! Outposts was cool and it gave people a solution but this is the ultimate hack😎

the real test is how it handles the weird edge cases in app store builds

Lovely!

@dabit3 Okay do you have mobile-use now? to test them any framework you use for Agents

lol you don't need Devin for this.

@imjaredz Can we set this up on our own Mac mini’s?

Hi! How it can test the Android on the side? Asking for a friend 📝

Damn what's the RAM on these VMs?

does devin have ios containers

The interesting unlock is not just device coverage but the feedback loop: an agent can now test the UI it changed instead of handing that last mile back to a human. How much of the validation is already automated in Devin?

can it run the smallest iphone and a pro max side by side too??

pretty cool @AnjanKatta