Percy Liang ✓
@percyliang · Stanford, CA · joined 31 Oct 2009
professor of computer science @Stanford @stanfordnlp, co-founder of @togethercompute, https://t.co/7R5THVnJ6u, @simile_ai, pianist
121 567Followers
424Following
1 337Posts total
620.7KViews on collected posts
Últimas publicaciones
Buscando publicaciones nuevas…
Marin 535B-A23B is 13% through training. This hero run would not be possible without the generous support of the Jen-Hsun and Lori Huang Foundation, which provided the funding for the compute (Coreweave). Thanks @JensenHuang for supporting open models! https://t.co/1qBdgV79S8
45.9K views · 749 likes
· 34 reposts · 15 replies
Open on X →
The "What-If Machine" is a vivid articulation of what Simile is building. It's not just about forecasting the future passively. It's about understanding how active interventions on the world change its course. It's the classic difference between causation and correlation.
16.2K views · 85 likes
· 5 reposts · 2 replies
Open on X →
Marin 535B-A23B is ~7% done training, and so far things look on track. Next Tuesday (Sept 1 @ 10 PT), we will have a Zoom panel/discussion where the Marin team will talk about the design decisions that went into this run, the tradeoffs made, and our learnings. If you're https://
82.1K views · 1.3K likes
· 80 reposts · 23 replies
Open on X →
The fact that Simile’s first technical blog post is about confidence is notable. Confidence is paramount to simulation. If a coding agent messes up, you can often tell and repair. If a simulation is wrong, you might never know and make a bad consequential decision based on it.
52.4K views · 391 likes
· 23 reposts · 8 replies
Open on X →
@percyliang congrats team! Excited that there's compute out there for non profits :) and will look forward to working with the model
12.7K views · 119 likes
· 0 reposts · 1 replies
Open on X →
Look at our data composition:
https://t.co/ihNKQh93rf
Watch the run live on wandb:
https://t.co/Bc3CcV4FOm
See all the gory details on GitHub:
https://t.co/swss50Hkbb
Assembling this hero run really required data, architecture, infra, kernels to all come together and was a huge
90.7K views · 648 likes
· 51 reposts · 6 replies
Open on X →
In addition to forecasting our final loss, we can also forecast the loss of intermediate checkpoints, so we can see if we’re on track over the voyage: https://t.co/CivuJfdPtT
43.9K views · 357 likes
· 10 reposts · 4 replies
Open on X →
@percyliang looks really great! one question - is anyone fulltime on Marin (aka, it's their fulltime job), or is it made up of people volunteering time?
2.6K views · 8 likes
· 0 reposts · 2 replies
Open on X →
Thanks to the Google TPU Research Cloud program for providing the much needed compute for this project. Thanks to the open-source community (including @AiEleuther @allen_ai @huggingface @llm360 @togethercompute @nvidia) for releasing tools and datasets. We are excited to give
4.6K views · 51 likes
· 0 reposts · 1 replies
Open on X →
To learn more about Marin, visit:
- Website: https://t.co/Kn8ynNMz7u
- GitHub: https://t.co/wPCG2KrIDl
- Discord: https://t.co/Lw7rHUmlVE
- Documentation: https://t.co/tCwe0WeBHN
- Announcement: https://t.co/qhXM9bmzSb
5K views · 41 likes
· 3 reposts · 1 replies
Open on X →
Have a specific use case? Come to our Datashop to curate data and train models.
Here’s how we curated more math data:
https://t.co/PH4Yv4q9Lt
Check out the data:
https://t.co/XHq7NC10OI https://t.co/zub34K8yjC
3.3K views · 33 likes
· 3 reposts · 1 replies
Open on X →
Have a new algorithm for training? Choose your compute budget and get on the speedrun leaderboard: how fast can you drive down validation loss?
https://t.co/1oGTrcHAv5 https://t.co/zTleS6rLRs
3.7K views · 40 likes
· 3 reposts · 1 replies
Open on X →
We’re also training Marin-32B-Base: it’s training - watch it live: https://t.co/cMFhErqBo2!
All models are Apache 2.0 and available at https://t.co/qu74DYj33v
4.2K views · 41 likes
· 3 reposts · 1 replies
Open on X →
We have trained some respectable models from scratch!
- Marin-8B-Base: beats Llama 3.1 8B on 14/19 benchmarks
- Marin-8B-Instruct: try it out on Together (https://t.co/OVy7Gr4gd7)
4.8K views · 45 likes
· 4 reposts · 1 replies
Open on X →
What would truly open-source AI look like? Not just open weights, open code/data, but *open development*, where the entire research and development process is public *and* anyone can contribute. We built Marin, an open lab, to fulfill this vision: https://t.co/racsvmhyA3
241.8K views · 1.4K likes
· 249 reposts · 72 replies
Open on X →
Marin (https://t.co/Kn8ynNMz7u) repurposes GitHub, which has been successful for open-source *software*, for AI:
1. Preregister an experiment as a GitHub issue
2. Submit a PR, which implements the experiment in code
3. PR is reviewed by experts in the community
4. Watch the https
6.8K views · 86 likes
· 7 reposts · 2 replies
Open on X →
Frente a cuentas del mismo tamaño
7 publicaciones de los últimos 90 días, junto al rango de 100K–1M seguidores. llega a más gente que cuentas de su tamaño.
Visualizaciones medianas45 851esta cuenta5 983mediana de 100K–1M
Alcance, %37.72%esta cuenta1.49%mediana de 100K–1M
Interacción, %0.85%esta cuenta0.93%mediana de 100K–1M
| Métrica | Esta cuenta | Mediana de 100K–1M | Proporción |
|---|---|---|---|
| Visualizaciones medianas por publicación | 45 851 | 5 983 | 7.66× |
| Alcance (visualizaciones ÷ seguidores) | 37.72% | 1.49% | 25.3× |
| Tasa de interacción | 0.85% | 0.93% | 0.91× |
Otras cuentas de este rango → Comparar con otra cuenta → Cómo se construyen estas referencias →
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
4.8K19 May
4.2K
3.7K
3.3K
5K
4.6K
2.6K
43.9K21 Aug
90.7K
12.7K
52.4K25 Aug
82.1K28 Aug
16.2K2 Sep
45.9K
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
1.04%19 May
1.07%
1.17%
1.11%
0.89%
1.13%
0.43%
0.85%21 Aug
0.79%
0.95%
0.81%25 Aug
1.75%28 Aug
0.58%2 Sep
1.76%
Reactions — likes, reposts, replies and quotes — divided by views. Median for 100K–1M accounts is 0.93%.
What the audience does
Likes66.0%5 398 in total
Reposts5.8%475 in total
Replies1.7%141 in total
Quotes1.0%80 in total
Bookmarks25.5%2 088 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.
Cuentas similares
Charles Gasparino ✓
@CGasparino
243.1K followers
· 125.8K posts
Kelly ✓
@kellyinvegas
243.1K followers
· 104.5K posts
Igor Assunção
@Igortep
243.1K followers
· 8.9K posts
Rhino.fi ✓
@rhinofi
243.1K followers
· 5.9K posts
Naga Vamsi ✓
@vamsi84
243.1K followers
· 3.1K posts
Diario de Sevilla
@diariosevilla
243.1K followers
· 390.3K posts
エカニス・エニカ🅾️ ✓
@EchanisEnicha
243.1K followers
· 1.2K posts
UNO
@realUNOgame
243.1K followers
· 27.5K posts