what people don't realize is that @luaroncrew and I met 5y ago, failed our first startup together
we joined again more than 1.5y ago to start working on http://boat.dev (before even the left of this chart)
we came into sandboxes as frustrated users, and we developed taste and expertise for them
slowly, all from 1st principles, all with a user-centric approach
we wanted to build the openclaw for software engineering, before openclaw, before long running cloud agents, before cursor had a cloud desktop, before the surge of ADEs
no sandbox was good enough for our usecase, and I believe now except http://boat.dev none would still be good enough
we wanted to build:
1. a always on CTO agent "Alfred" that would take high level goals & constantly loop (before loops) on your stuff:
this needed a always on sandbox, still cheap if running 24/7 but beefy as f, with ability to setup an entire dev environment + browser for acquiring info or signal to loop on + a desktop to record demos on to keep deliverables super high level
2. the ability for the CTO agent to make copies of his entire disk and setup live into shorter running sandboxes with same OS and all, just for individual subagents and tasks that needed isolated testing
this needed a robust fs snapshot system, git-like in the background, optimized for small snapshot size, yet robust and predictable
all of this at a time of still great agent flackyness
the sandbox got easy to use, (cost) optimized and unique enough that we thought maybe it'd have more success as a standalone product
and we were so go damn right, our ARR has 34x since we shipped http://boat.dev in late may
so yeah, we had to grind for monthes to get something okay-ish
overnight successes don't exist, great products require full honesty, full dedication, full grind, and knowing the pain as good as having lived it first hand