Nathan Lambert ✓
@natolambert · Seattle · joined 24 Dec 2014
Open model research @ something new. Prev. co-led Olmo at Ai2. Writes @interconnectsai, wrote https://t.co/alRXKINTwE
101 777Followers
941Following
12 063Posts total
3.3MViews on collected posts
Последние посты
Ищем новые посты…
OOO for a bit, time to celebrate & rest :) https://t.co/SUDnav5fYE
17.2K views · 255 likes
· 1 reposts · 15 replies
Open on X →
In the frantic era feeling early RSI effects, sending friends physical books, hosting events, chatting with students early on their ai journey, etc are all wonderful to do. Many things AGI won’t replace, and it’s easy to forget to do them. I’ll visit a few more universities!
10.4K views · 86 likes
· 6 reposts · 12 replies
Open on X →
just received a great bed time reading book by the mail looking forward for tonight https://t.co/aKoXDIVVV3
19.3K views · 392 likes
· 23 reposts · 7 replies
Open on X →
Over the weekend I had Codex parse 500K arXiv AI/ML papers since ChatGPT to understand which open models are used for research.
In 2024, ~30% of papers mentioned an American open model and only 10% a Chinese model.
Today, ~40% of papers mention a Chinese (open) LLM, and only h
94.3K views · 449 likes
· 63 reposts · 45 replies
Open on X →
So happy for the whole Marin team. They've been quietly working up to this for a long time. I'm excited to get to work with the community post-training this model!
44.1K views · 436 likes
· 12 reposts · 8 replies
Open on X →
@natolambert Thanks! I agree with 1-5. I find 6 hard to define precisely, but generally makes sense.
I have no idea whether 1-5 are a "sufficient" explanation, how much of a gap there is left to explain, after accounting for idea diffusion / 4 minute mile effects.
My current be
4K views · 4 likes
· 0 reposts · 1 replies
Open on X →
@natolambert Kimi k3>>>GLM 5.3 overall from my experience, especially at coding and research tasks.
1.6K views · 9 likes
· 0 reposts · 0 replies
Open on X →
@natolambert honestly refreshing to see someone admit the denial part instead of pretending they called it early
959 views · 2 likes
· 0 reposts · 0 replies
Open on X →
GLM 5.3 notes and why we should stop being so surprised about these very strong Chinese models (most of this is talking myself through some of my denial -- yes, these models are the real deal).
https://t.co/hePI3fILRF
120.4K views · 491 likes
· 38 reposts · 16 replies
Open on X →
@natolambert i thought you were joking but this is actually what happened…? wtf
19.3K views · 216 likes
· 1 reposts · 1 replies
Open on X →
@natolambert impressive where next word prediction gets you, this was probably in the training data
23.2K views · 453 likes
· 1 reposts · 7 replies
Open on X →
TLDR: An openai model, during evaluation on a cyber benchmark, exploited a public zero day bug, escaped sandboxing in openai's infra, and got into the internal huggingface infra via an exploit (through a public dataset service) all in the attempt to solve a benchmark problem.
925.3K views · 7.4K likes
· 471 reposts · 180 replies
Open on X →
@natolambert Big congrats Nathan! RLWF next? :) https://t.co/hbv2IxGntu
1.4K views · 7 likes
· 0 reposts · 0 replies
Open on X →
The people who did a mix of directly helping me on the book, inspiring me, and or helping the book without knowing it!
Luca Soldaini: @soldni
Kyle Lo: @kylelostat
Dirk Groeneveld: @mechanicaldirk
Hanna Hajishirzi: @HannaHajishirzi
Saumya Malik: @saumyamalik44
Costa Huang:
8.8K views · 60 likes
· 3 reposts · 2 replies
Open on X →
Website: https://t.co/Btt5QL56oV
Buy on Amazon: https://t.co/MY5eyxGkpt
Buy on Manning: https://t.co/Rj3xWxNq4Q
Course landing page: https://t.co/P0Doi5SCUx
Course Playlist: https://t.co/C1lJjmXPMG
9.8K views · 151 likes
· 12 reposts · 10 replies
Open on X →
My book, Reinforcement Learning from Human Feedback is done!
This is the book I wish I had when learning to fine-tune, align, & now post-train models since ChatGPT. The resource has been built by me finding time to study and document the fundamentals on nights and weekends since
264.3K views · 3.3K likes
· 273 reposts · 151 replies
Open on X →
Excited to co-organize the 1st "RL from World Feedback" workshop at ICML on July 10, featuring:
- @MillionInt : ex VP of RL @ OpenAI
- @chelseabfinn : Stanford/PI
- Ben Eysenbach: Princeton
- @brianzhan1 : Partner @ https://t.co/ZiTf1kOkHj
- @robertarail : GDM
- Jesse Zhang: UW h
42.6K views · 196 likes
· 18 reposts · 2 replies
Open on X →
And what this means: https://t.co/6P4qPkupV7
6.1K views · 23 likes
· 2 reposts · 2 replies
Open on X →
@natolambert Shadow-banning LLM development was certainly not on my bucket list for 2026
8.9K views · 249 likes
· 9 reposts · 6 replies
Open on X →
Also, what about building LLMs for safety or science applications or education or whatever. Gross
12.4K views · 228 likes
· 3 reposts · 5 replies
Open on X →
I don't want them to do this but it's totally in their right to do so. Makes the open frontier obviously more strategically valuable.
10.5K views · 218 likes
· 4 reposts · 10 replies
Open on X →
Labs starting to pull up the ladders on the ability to diffuse AI was inevitable. Doing it without telling the user is misaligned.
295.8K views · 1.9K likes
· 183 reposts · 56 replies
Open on X →
When Fable 5 is used for frontier LLM development, it does not notify the user and instead limits the model’s capabilities through methods such as prompt modification, steering vectors, and PEFT.
Anthropic estimated that this would affect approximately 0.03% of traffic. https://
1.3M views · 1.4K likes
· 154 reposts · 96 replies
Open on X →
На фоне аккаунтов своего размера
9 постов за последние 90 дней рядом с диапазоном 100K–1M подписчиков. показывается широко, но откликается мало кто.
Медианные просмотры17 228этот аккаунт5 983медиана для 100K–1M
Охват, %16.93%этот аккаунт1.49%медиана для 100K–1M
Вовлечённость, %0.60%этот аккаунт0.93%медиана для 100K–1M
| Показатель | Этот аккаунт | Медиана для 100K–1M | Отношение |
|---|---|---|---|
| Медианные просмотры на пост | 17 228 | 5 983 | 2.88× |
| Охват (просмотры ÷ подписчики) | 16.93% | 1.49% | 11.4× |
| Вовлечённость | 0.60% | 0.93% | 0.65× |
Другие в этом диапазоне → Сравнить с другим аккаунтом → Как считаются эти ориентиры →
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
1.4K21 Jul
925.3K
23.2K
19.3K
25.9K
120.4K14 Aug
959
1.6K
4K
44.1K21 Aug
94.3K24 Aug
19.3K
10.4K25 Aug
17.2K
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
0.49%21 Jul
0.88%
1.98%
1.13%
2.76%
0.46%14 Aug
0.21%
0.55%
0.13%
1.03%21 Aug
0.60%24 Aug
2.20%
1.00%25 Aug
1.58%
Reactions — likes, reposts, replies and quotes — divided by views. Median for 100K–1M accounts is 0.93%.
What the audience does
Likes72.0%18 560 in total
Reposts5.0%1 295 in total
Replies2.5%632 in total
Quotes1.9%490 in total
Bookmarks18.6%4 798 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.
Похожие аккаунты
NATE ✓
@NATERERUN
203.5K followers
· 26.7K posts
Jennifer Arnold, MD
@JenArnoldMD
203.5K followers
· 17.2K posts
Whale Taylor🌸🐋ホエール・テイラー ✓
@whaletaylor
203.5K followers
· 7.2K posts
Columbus Dispatch
@DispatchAlerts
203.5K followers
· 262K posts
The Random Guy ✓
@RandomTheGuy_
203.5K followers
· 9.3K posts
あらなるめい公式
@aranarumey_info
203.5K followers
· 639 posts
Família Te Atualizei ✓
@newsteatualizei
203.5K followers
· 2.8K posts
早瀬ノエル（FRUITS ZIPPER） ✓
@noel_fz1229
203.5K followers
· 14.4K posts