tweetindex
TR

Karan Dalal

@karansdalal

6 574Followers
382Following
100Posts total
599.6KViews on collected posts

Son gönderiler

@karansdalal great work! you should release some weights so we can do experiments :) it was really eye-opening to tinker with mamba when it first came out, both in strengths/weaknesses and also straight up totally different emergent behavior with out of domain tasks. 7.9K views · 17 likes · 0 reposts · 2 replies Open on X →
@karansdalal This will be one of the important themes for this year. Loved the paper. My piece on it with comments from authors: https://t.co/uf8AK9PSbf 4.2K views · 22 likes · 2 reposts · 2 replies Open on X →
@karansdalal Adding this to my reading list this week. Thanks for sharing! 6.5K views · 15 likes · 0 reposts · 1 replies Open on X →
LLM memory is considered one of the hardest problems in AI. All we have today are endless hacks and workarounds. But the root solution has always been right in front of us. Next-token prediction is already an effective compressor. We don’t need a radical new architecture. The h
581K views · 2.1K likes · 318 reposts · 92 replies Open on X →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

581K12 Jan
6.5K
4.2K
7.9K13 Jan

Last 4 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.44%12 Jan
0.25%
0.62%
0.24%13 Jan

Reactions — likes, reposts, replies and quotes — divided by views.

What the audience does

Likes47.9%2 116 in total
Reposts7.2%320 in total
Replies2.2%97 in total
Quotes1.7%76 in total
Bookmarks40.9%1 809 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.

Benzer hesaplar