Stanislas Polu ✓

@spolu · Paris · joined 26 Nov 2007

co-founder+engineer(https://t.co/SXBR0l9TrF); alumni(https://t.co/z6zJ8xaKGI, https://t.co/CvVTA1CHAo, https://t.co/WOVEe2aLcK, https://t.co/ui9I4Nj7o1);

28 197Followers

659Following

9 075Posts total

284.5KViews on collected posts

Posts recentes

Stanislas Polu ✓@spolu9 Jul

@spolu oooo we should collab. ;D https://t.co/eB1nA3PElf similaritieeeessss 104 views · 1 likes · 0 reposts · 0 replies Open on X →

Stanislas Polu ✓@spolu8 Jul

@spolu The Skill suggestions queue feels like the right primitive here. Each suggestion needs the raw interaction, the extracted rule, and a before/after eval case, or the company brain can absorb a plausible edit with no way to reproduce why it mattered. 146 views · 1 likes · 0 reposts · 0 replies Open on X →

Stanislas Polu ✓@spolu8 Jul

@spolu the batch step is doing more than it looks. when two interactions disagree, the merge picks a winner before the owner ever sees the diff. so they approve the final text, not the fact that there was a conflict underneath it. does the owner get to see an edit came from a 148 views · 1 likes · 0 reposts · 0 replies Open on X →

Stanislas Polu ✓@spolu8 Jul

Self-improving skills are now available on Dust. A new important step in the engineering of our harness towards recursive self improvement. Skills are highly curated capabilities (tools + knowledge + instructions + code) reusable across agents. They are the brain of your https: 260K views · 84 likes · 7 reposts · 12 replies Open on X →

Stanislas Polu ✓@spolu7 Jul

Fascinating read from A\: [0]. The method to compute the Jacobian Lens is described here[1]. Interestingly the J-lens is one `dxd` matrix per layer and using it on an activation is a simple operation. The lens is the average on pre-training data of the impact of a perturbation 2.6K views · 24 likes · 3 reposts · 1 replies Open on X →

Stanislas Polu ✓@spolu3 Jul

We've been also tracking % messages posted in channels for years now. Same motivation since inception in 2023: The quality of your AI agents is directly proportional to the quality of your company's shared context. DMs, undocumented meetings, hallway conversations… these are ht 3.7K views · 16 likes · 3 reposts · 0 replies Open on X →

Stanislas Polu ✓@spolu30 Jun

Great chat with @collinmathilde at YC Startup School Paris yesterday. Loved the energy in the room as well as the weird basement backstage consisting of a very small window-less room packed with founders😁 I guess that’s what they meant by “founder mode” on the caps they offered h 2.7K views · 30 likes · 1 reposts · 2 replies Open on X →

Stanislas Polu ✓@spolu22 May

Agents need servers not computers: The emergence of hyper-personalized software. Why localhost is dead and why it's all exciting. https://t.co/16jNTq7UcM 15.1K views · 30 likes · 3 reposts · 3 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

15.1K22 May

2.7K30 Jun

3.7K3 Jul

2.6K7 Jul

260K8 Jul

148

146

1049 Jul

Last 8 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.24%22 May

1.20%30 Jun

0.51%3 Jul

1.10%7 Jul

0.04%8 Jul

0.68%

0.96%9 Jul

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

What the audience does

Likes54.4%187 in total

Reposts4.9%17 in total

Replies5.2%18 in total

Bookmarks35.5%122 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.

Posts recentes

Growth & engagement

Views per post

Engagement rate per post

What the audience does

Contas semelhantes