Maybe the DO or Paperspace team can build something like RunPod. They support containers and per-minute billing.
Or like replicate.com but find a way to have a model sleep without being completely unloaded.
Maybe a good business model could be a system built around LoRAs or the equivalent for LLMs. That way you could share a lot of instances running the same containers but just swap out the adaptation data quickly for different customers.
paper space has such an oddball product - they first started off as some sort of cloud Virtual Desktop type of thing and this seems to be their new pivot.
DigitalOcean made some missteps recently with massive cuts to their community team (technical writing group)
> DigitalOcean made some missteps recently with massive cuts to their community team (technical writing group)
This team, as a consumer, was top-notch. That being said, it's hard to make content-based justify its cost. If DO revenues were 2-3X what they are now, I bet they'd keep this great team (from an output perspective) in tact.
5 comments
[ 3.2 ms ] story [ 31.7 ms ] threadOr like replicate.com but find a way to have a model sleep without being completely unloaded.
Maybe a good business model could be a system built around LoRAs or the equivalent for LLMs. That way you could share a lot of instances running the same containers but just swap out the adaptation data quickly for different customers.
DigitalOcean made some missteps recently with massive cuts to their community team (technical writing group)
This team, as a consumer, was top-notch. That being said, it's hard to make content-based justify its cost. If DO revenues were 2-3X what they are now, I bet they'd keep this great team (from an output perspective) in tact.