Robert Nishihara ✓

@robertnishihara · joined 19 Mar 2009

Co-founder @anyscalecompute. Co-creator of @raydistributed. Previously AI research at Berkeley.

17 182Followers

872Following

1 829Posts total

619.6KViews on collected posts

Against accounts of the same size

7 posts from the last 90 days, next to the 10K–100K follower range. shown widely, but few of those viewers react.

Median views22 635this account2 751median for 10K–100K

Reach, %131.74%this account9.00%median for 10K–100K

Engagement, %0.93%this account1.65%median for 10K–100K

Metric	This account	Median for 10K–100K	Ratio
Median views per post	22 635	2 751	8.23×
Reach (views ÷ followers)	131.74%	9.00%	14.6×
Engagement rate	0.93%	1.65%	0.56×

Others in this range → Compare with another account → How these benchmarks are built →

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

66.3K1 Jun

166.3K15 Jun

22.6K21 Aug

1.8K

2.5K26 Aug

187.4K

38.4K27 Aug

129.2K1 Sep

5.1K

Last 9 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.11%1 Jun

0.14%15 Jun

0.93%21 Aug

0.94%

1.24%26 Aug

0.19%

0.28%27 Aug

0.65%1 Sep

0.96%

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.65%.

What the audience does

Likes71.2%1 655 in total

Reposts7.5%174 in total

Replies3.0%70 in total

Quotes1.8%41 in total

Bookmarks16.5%383 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.

Latest posts

An agentic RL training guide for Qwen 3.5 397B with SkyRL. They show very large improvements with roughly 2000 expert-labeled tasks. 5.1K views · 42 likes · 4 reposts · 3 replies 01 Sep 2026 How does one RL post-train a 397B model for long-horizon knowledge work? 👩‍💼 We share every step we took to bring Qwen 3.5 397B from 16.1% Pass@1 to 27.3% on APEX-Agents using DPPO, including final models weights and the full training script🚀 This is the first of many works from 129.2K views · 709 likes · 100 reposts · 16 replies 01 Sep 2026 I got some as well. Inspirational talk! Also, very cool to see that @periodiclabs is using Kimi to reduce inference costs 20-50x over frontier APIs. Everyone will do this. https://t.co/UtbF8O2Fy6 38.4K views · 98 likes · 2 reposts · 6 replies 27 Aug 2026 Does your boyfriend give keynotes and look this good? MINE DOES!!!! @LiamFedus https://t.co/I4Fit85s6n 187.4K views · 326 likes · 8 reposts · 24 replies 26 Aug 2026 Really enjoyed this talk from @AndrewLBeam at @LilaSciences at Ray Summit today on the path toward automating scientific research: intelligence, software, & robotics. https://t.co/wbatM0XtAT 2.5K views · 23 likes · 4 reposts · 3 replies 26 Aug 2026 This is next week in SF! 1.8K views · 16 likes · 0 reposts · 1 replies 21 Aug 2026 vLLM Conference is next week, and we have a packed schedule 🎊 📅. Here's the full list of events you should know: Mon: 🔷 4–6PM vLLM × Ray × Google Cloud Happy Hour: https://t.co/hc0IxmdUkI 🔷 6–9PM vLLM × Dynamo Meetup: https://t.co/OWQ6eMij0d Tue: 🔷 11–11:30AM vLLM https://t. 22.6K views · 170 likes · 31 reposts · 5 replies 21 Aug 2026 Beyond pre-training, here's how I imagine most learning will work. 1. AI models / systems will maintain large collections of retrievable knowledge. This will include facts like "the capital of California is Sacramento" and tactics like "when playing Monopoly, buy a bunch of 166.3K views · 208 likes · 20 reposts · 10 replies 15 Jun 2025 We're missing techniques for "training-time reasoning." Right now there's a lot of progress on inference-time reasoning, which is incredibly cool (I use o3 all the time). If I think about how I learn stuff, e.g., when reading a technical paper, it's very compute intensive. Most 66.3K views · 63 likes · 5 reposts · 2 replies 01 Jun 2025