I'm thinking Twitter did something to its UI that drops the XBlock inference sensitivity dramatically, obvious Twitter screenshots are coming back as misses with like 75% certainty (threshold is 95%). I'm going to put extra weight on new Twitter screenshots next training run to compensate.
Hey check out my cool new website https://shimeji.us-east.host.bsky.network/xrpc/com.atproto.sync.getBlob?did=did:plc:kkf4naxqmweop7dv4l2iqqf5&cid=bafkreibul76fnm2jdkvjnv7mscax5macltoo2jnqfasnge3y7gzqw2nive
Dropped an idea in Bluesky's GitHub discussions about verification. I admit I'm not particularly expert in OAuth or SSO so I very much welcome feedback, please let me know if I'm missing something obvious with this idea. [github.com/bluesky-soci...]( ) [How to confirm ownership of a ...]( )
The fact that this is an open platform where you can try wild moonshot ideas like this (I'm not on the Bluesky team, I'm just a simple gal who likes emoji and hates Twitter screenshots) is a hugely underrated aspect of this place. Can you imagine trying to build something like this on *Threads*? 🙅🏼‍♀️ RE: View quoted note →