@philzona Simplicity. Fly does everything I need out of the box with code. I have a cluster in Vultr on standby for advanced k8s labs though. Obviously, if I hit "scale", the math changes.
@philzona This is my problem! For hosting lab sandboxes, this is an amazing time, but the timeouts kill like 90% of my options because I'm using them outside of their use case. I obviously don't complain because I know I'm a weirdo.
GitHub: there are WAY too many manual things I have to do in the UI to get things done. I honestly don't know how you're getting DDoSed like this when it's this hard to auth anything.
@liran_tal I built on the Strava API because it was easiest, but I do subscribe to https://t.co/zZrGxjMmip also. It's a fantastic platform. I paused work on the tool while I heal from a foot injury, but I'll be getting back to it soon.
No, your open model is not as good as Opus. Period. End of story. The "AI is not good" sentiment is largely coming from the delusion that an open model is capable of doing more than extremely focused tasks without spending unfathomable amounts of money.
@liran_tal@typesafeai it's incredibly small and runs on a lot of camera hardware, raspberry pi, etc. No hefty hardware needed. There might be a hosted model out there also.