Experimental browser for the Atmosphere
Loading post...
{ "uri": "at://did:plc:ygzrpn6mqpibaf6rowkd3uh6/app.bsky.feed.like/3lot7wdvsoy2g", "cid": "bafyreig6xsxjozwptromsv3l2fi4caucytbdsbsupkzbshmv24ljexfvuy", "value": { "$type": "app.bsky.feed.like", "subject": { "cid": "bafyreidxaa22u4cvjx5k73wzbmcznodsj5meeedwrgoioqs46nqfm7jj34", "uri": "at://did:plc:vg3thtvfbgfrr3u6pf6hy3yk/app.bsky.feed.post/3lot6qkuzsk2z" }, "createdAt": "2025-05-10T15:09:36.550Z" } }
Further cogitation: Chinchilla scaling laws were based on loss analysis, so base model ability to predict next token. This is a very different target objective as using language as a proxy for logical relations (in effect "reasoning" models).
May 10, 2025, 2:48 PM