Zhengyao Jiang ✓

@zhengyaojiang · London · joined 31 Oct 2015

Cofounder & CEO @WecoAI - automated hill climbing with LLMs. Prev: PhD in ML @UCL_DARK. (Zheng=j-uhng, j as in job; yao=y-aoww)

13 400Followers

741Following

885Posts total

4MViews on collected posts

Latest posts

Zhengyao Jiang ✓@zhengyaojiang12 Sep

As people delegate more research and engineering tasks directly to agents, I think the gap between achieving a goal and understanding how it was achieved will become a growing problem. I’m a bit torn about this, as someone working on autoresearch who was also trained as a 1K views · 11 likes · 2 reposts · 7 replies Open on X →

Zhengyao Jiang ✓@zhengyaojiang11 Sep

A great read. I have similar feelings about how AI labs approach progress directly and without nurturing of scientific communities & intuition. The math research community went through the transition the fastest, so it was felt most. Other fields next. https://t.co/M2uCXIERyC 39.4K views · 551 likes · 66 reposts · 26 replies Open on X →

Zhengyao Jiang ✓@zhengyaojiang8 Sep

With autonomous research, it probably won't take a decade before most frontier research is beyond the understanding of the best human experts. This feels like sci-fi. We'll live in a super fast progressing world shaped by discoveries that only AI can fully understand 604 views · 5 likes · 2 reposts · 2 replies Open on X →

Zhengyao Jiang ✓@zhengyaojiang31 Aug

I'm always a bit surprised this thread of work keeps making progress. How come completely different forms of time series can be modeled by a single model, and there's even positive transfer? Almost feels like a deeper mystery than LLMs 103.5K views · 1.1K likes · 54 reposts · 33 replies Open on X →

Zhengyao Jiang ✓@zhengyaojiang31 Aug

Introducing TimesFM-3, a state-of-the-art time series foundation model that enables accurate multivariate time series forecasting in a single forward pass, significantly outperforming other forecasting models across major benchmarks. More on the blog →https://t.co/uSlnIdUJ4Q htt

2M views · 11.6K likes · 1.3K reposts · 185 replies Open on X →

Zhengyao Jiang ✓@zhengyaojiang14 Aug

Had a look at Cordis, the theory behind DeepSeek Harness's plugin architecture. Quick thoughts: It's basically a theory that, as long as you follow a set of contracts for the components of a harness, there are theoretical guarantees that your harness can safely remove a 36K views · 407 likes · 25 reposts · 22 replies Open on X →

Zhengyao Jiang ✓@zhengyaojiang14 Jul

The first experimental evidence of recursive self-improvement (RSI). Autoresearching the autoresearch agent for eight days. The result beats the harness we hand-tuned for two years, on held-out benchmarks: 🧵(1/7) https://t.co/8kfmR2AlT7

0:40

1.8M views · 9.6K likes · 1.2K reposts · 266 replies Open on X →

Against accounts of the same size

7 posts from the last 90 days, next to the 10K–100K follower range. shown to more people than peers of the same size.

Median views39 389this account924median for 10K–100K

Reach, %293.95%this account3.62%median for 10K–100K

Engagement, %1.26%this account1.52%median for 10K–100K

Metric	This account	Median for 10K–100K	Ratio
Median views per post	39 389	924	42.6×
Reach (views ÷ followers)	2.9× audience	3.62%	81.2×
Engagement rate	1.26%	1.52%	0.83×

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

1.8M14 Jul

36K14 Aug

2M31 Aug

103.5K

6048 Sep

39.4K11 Sep

1K12 Sep

Last 7 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.61%14 Jul

1.26%14 Aug

0.65%31 Aug

1.10%

1.49%8 Sep

1.63%11 Sep

1.97%12 Sep

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.

What the audience does

Likes47.9%23 223 in total

Reposts5.4%2 625 in total

Replies1.1%541 in total

Quotes0.4%211 in total

Bookmarks45.2%21 900 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.