🎉 Launching online today:
"Calculating Empires: A Genealogy of Technology and Power since 1500."
Dive into our large-scale interactive visualization exploring how technical and social structures co-evolved over five centuries. 🧵 https://t.co/Z6sfaCevDf
We think a great use-case for this is finding folks on repos that would be used to roll whatever you're building internally. Then, you know they are already trying to solve the problem!
Just be respectful and maybe don't scrape all your competitors' users :)
Happy stargazing!
... because GitHub's API only lets you return 100 items at a time, we need to make <num_stars>/100 requests for each repo. That can add up quickly! Not only that, but since we use GQL, and their pagination cursor's aren't just page indices... we have to do it all one at atime.
If you know a better way of doing this please tell us so we can make this wayyy faster because right now it takes forever on large repos, and repos with >10k stars are horrifically slow. Maybe caching on our side could help a bit?
GitHub's rate-limits are pretty restrictive without being signed in (100 reqs/hr). So, we let you optionally sign in (5k/hr). All requests happen client-side, which you can verify in the network tab, so we never see your token. What we can do is chew through 5k reqs... slowly...
https://t.co/j163mYUJTS is a webapp to help you analyze stargazers in GitHub repos. It's really simple - just paste in a repo, filter for users matching the properties you want (eg. an email, company, etc.), and copy/download the CSV!
Ever wanted to email folks who starred a GH repo relevant to your product? Us too! We couldn't find any good ones for free or cheap so... we spent a couple hours building it last night!
Everyone is asking for 500M on 15T tokens but do we even know what the limit is for 1M-10M models? Surely they are under-parameterized, but what are the emergent capabilities (if any) for a 10M model @ 1T vs 10T?
@som1twisted I'm less worried about the next attack and more worried about existing backdoors tbh. Well-executed supply chain attacks like this reveal a much deeper problem about open-source maintainers and burn out imo