Han Xiao ✓

@hxiao · Sunnyvale, CA · joined 05 Apr 2009

VP, AI @Elastic prev: founder & ceo @JinaAI_

20 793Followers

280Following

6 065Posts total

27.4KViews on collected posts

Latest posts

Han Xiao ✓@hxiao5 Sep

This is actually Nevada. https://t.co/DlgyoVIYHb 220 views · 1 likes · 0 reposts · 0 replies Open on X →

Han Xiao ✓@hxiao5 Sep

Fallen leaf lake https://t.co/fVYQRxLEIU 521 views · 7 likes · 0 reposts · 1 replies Open on X →

Han Xiao ✓@hxiao4 Sep

Sugar pine point / Lake Tahoe https://t.co/5phYTeJeq6 1.1K views · 12 likes · 0 reposts · 1 replies Open on X →

Han Xiao ✓@hxiao4 Sep

@hxiao I just switched from @claude_code to @PrimeIntellect's prime agent which is built with RLM in mind. My session has been running for 50h with Qwen3.8 Flash and the token usage is way better than what CC was doing. CC outputted ~4.5M token while processing 360M vs ~2.3M fo 460 views · 3 likes · 0 reposts · 1 replies Open on X →

Han Xiao ✓@hxiao3 Sep

The second is 24h & 9 laps. Same pi harness, but the model is qwen-3.8-27b-nvfp4 by @radixark on a ½ RTX6K on GCP CloudRun. 9 laps with the longest ran >14h without steering or nudging. The swimlane looks less clean than the 1st tho, bc the run got interrupted a couple of times h 1.4K views · 9 likes · 1 reposts · 1 replies Open on X →

Han Xiao ✓@hxiao3 Sep

anyway, I think this is a pretty fun experiment: a reality-check of pi + self-hosted models to run long-horizon tasks. Overall it's usable, and I'm thinking about offloading some my less competitive chores to this stack. An individual with unmetered intelligence can unlock a lot 809 views · 3 likes · 0 reposts · 0 replies Open on X →

Han Xiao ✓@hxiao3 Sep

Sharing two long-horizon sessions on pi + self-hosted models. first: 16 hours long, zero steering, 99 laps. I used pi + local models to port our upcoming Jina model from bf16 PyTorch into MLX+MTP quants (model J in the sequel). The local model is qwen3.8-flash-next-omlx-q4 from h 20K views · 310 likes · 14 reposts · 14 replies Open on X →

Han Xiao ✓@hxiao2 Sep

thought it's hallucinating, but turns out to be true, they did release qwen-3.7-text-embedding but seems api only https://t.co/HgZQ4r0Str 3K views · 13 likes · 1 reposts · 2 replies Open on X →

Against accounts of the same size

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

Median views948this account966median for 10K–100K

Reach, %4.56%this account3.66%median for 10K–100K

Engagement, %0.84%this account1.51%median for 10K–100K

Metric	This account	Median for 10K–100K	Ratio
Median views per post	948	966	0.98×
Reach (views ÷ followers)	4.56%	3.66%	1.25×
Engagement rate	0.84%	1.51%	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

3K2 Sep

20K3 Sep

809

1.4K

4604 Sep

1.1K

5215 Sep

220

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

0.54%2 Sep

1.70%3 Sep

0.37%

0.81%

0.87%4 Sep

1.20%

1.54%5 Sep

0.45%

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.51%.

What the audience does

Likes50.1%358 in total

Reposts2.2%16 in total

Replies2.8%20 in total

Quotes0.3%2 in total

Bookmarks44.6%319 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

Against accounts of the same size

Growth & engagement

Views per post

Engagement rate per post

What the audience does

Similar accounts