tweetindex
EN

Standard Intelligence ✓

@si_pbc · joined 08 Sep 2024

10 409Followers
0Following
89Posts total
1.6MViews on collected posts

Latest posts

Back when we were raising our seed round, Lachy was one of the only people in Silicon Valley who saw our idea, immediately got it, and wrote the check that let us train FDM-1. Incredibly grateful to have him as an early supporter. 13.9K views · 112 likes · 2 reposts · 1 replies Open on X →
@si_pbc is this firecracker based? 265 views · 1 likes · 0 reposts · 1 replies Open on X →
We built eval infrastructure that drives over 1M rollouts per hour across 80,000 forking VMs, each with 1 vCPU and 8GB of RAM; a single H100 can control 42 of these in parallel. These forking VMs are powered by @modal GPUs - the combination of instant spinup and low latency to 43.5K views · 297 likes · 15 reposts · 7 replies Open on X →
We've really enjoyed working with Mithril; spot & market compute is the way forward, and they've been an amazing partner. 4.1K views · 35 likes · 2 reposts · 0 replies Open on X →
Huge congratulations to our partners at Standard Intelligence! Thank you for using us as a compute partner. 6.4K views · 31 likes · 1 reposts · 0 replies Open on X →
@si_pbc truly everything is computer 3.7K views · 81 likes · 0 reposts · 3 replies Open on X →
True computer use is fully general. FDM-1 uses arrow keys on a computer to steer a car in San Francisco with less than 1 hour of fine-tuning data. The action policy is critical: tuning FDM-1 to drive gets much higher accuracy than tuning just the video encoder on the same data. 72.3K views · 412 likes · 16 reposts · 10 replies Open on X →
Here’s the blog post, where you can learn more about how we trained this model: https://t.co/lsVT0va26n 30.6K views · 294 likes · 12 reposts · 8 replies Open on X →
FDM-1 completes complex tasks and navigates interfaces well enough to use CAD applications. Forking VMs allow us to snapshot when a successful operation completes (extrusion, selection, etc.), letting us apply test-time compute to computer use. https://t.co/IjqKKf9u2I 59.9K views · 329 likes · 12 reposts · 4 replies Open on X →
We’ve made two main advances: the ability to train on our 11M+ hour computer action dataset and understand long-context video. Our video encoder can fit nearly two hours of 30FPS, high-resolution video into a 1M token context window, ~50x more efficient than existing SOTA. https 127K views · 684 likes · 24 reposts · 7 replies Open on X →
Inspired by VPT (Baker et al), we train an inverse dynamics model (IDM) to predict frame-by-frame computer actions. The IDM leverages 40k hours of contractor data to label 11 million hours of screen recordings—550,000x larger than the largest open-source computer use dataset. 40.5K views · 334 likes · 2 reposts · 1 replies Open on X →
Computer use models shouldn't learn from screenshots. We built a new foundation model that learns from video like humans do. FDM-1 can construct a gear in Blender, find software bugs, and even drive a real car through San Francisco using arrow keys. https://t.co/kepCzAKY3U 1.2M views · 3.9K likes · 394 reposts · 186 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

1.2M23 Feb
40.5K
127K
59.9K
30.6K
72.3K
3.7K
6.4K
4.1K
43.5K
26524 Feb
13.9K30 Apr

Last 12 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.39%23 Feb
0.84%
0.58%
0.59%
1.04%
0.63%
2.29%
0.52%
0.90%
0.74%
0.75%24 Feb
0.83%30 Apr

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

What the audience does

Likes60.7%6 538 in total
Reposts4.5%480 in total
Replies2.1%228 in total
Quotes2.8%303 in total
Bookmarks29.9%3 215 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.

Similar accounts