@rrhoover Thanks for the question Ryan, and for featuring us on Product Hunt! We first started Spell to solve the basic infrastructure problems that were preventing a lot of developers from getting started with deep learning. Having to either buy a GPU and build a rig or manage your own cloud infrastructure was making it unnecessarily hard, so we built the Spell CLI to make running code in the cloud in a safe and efficient way as simple as typing "spell run". The most common use case is a user will find an open-source implementation of a model from the research world or on GitHub and use Spell to run it themselves. This is the kind of tinkering that helps developers add DL to their toolkit, but we also see that most new research or industrial applications start from a reference implementation from the open source community. Specifically, we see a lot of LSTMs, Transformer networks and other NLP models applied to custom data, and a lot of people training GANs. This most recent product launch is focused on collaboration and security, and comes from a lot of time spent with companies doing deep learning for real-world applications. Getting these models to work on a new task is not trivial, and requires lots of experimentation and teamwork. So we built shared clusters that teams can bring online and allow them to share results, progress, notes etc. Please give it a try and let us know how you like it! Spell is free for individuals, and we charge GPU time at cost. Also, use the promo code PRODUCTHUNT and get $10 in free credits to use for compute! Add the promo code under Settings > Billing. Happy hunting, everyone, and let me know if you have any questions!