huangserva Verified by creator
@servasyy_ai X
Banned at 13.9K, he restarted and hit 10K again in 37 days
A veteran programmer and serial founder by his own account. His new account reposted the old one’s prompts and, he says, was back to 5K in 10 days. Then came long Agent and Codex tutorials; one Codex piece drew 566K.
- Report
- v1
- Data as of
- 2026-11-08
- Sample
- 342 days of posting counts
huangserva@servasyy_ai突然觉得身体疲乏,扛不住 (Suddenly feeling worn out, can’t keep going)
View on X →At a glance
- What it is
- A Chinese-language X account of long hands-on articles on Agents, Codex and local deployment.
- Who reads
- Developers using Claude Code, Codex, OpenClaw and local models.
- How it earns
- Sponsored posts for AI products (self-reported).
- How it broke out
- After his big account was banned he started over, won back old readers with its prompts, and was back to 10K in 37 days (self-reported).
X followers Fact
42.9K
2026-11-08
Back to 10K Self-reported
37 days
2025-12-11 → 2026-01-17
Median views, first two weeks Fact
6,745
38 main posts, 17 over 10,000
Replies / main posts Derived
4.5×
6,059 replies, 1,349 main posts
Summary
The play
After the ban he opened a new account at once and reposted the old account’s prompts and image-generation tricks to win back old readers. Then he wrote long hands-on tutorials on whichever Agent tool was hottest, with more than 4× as many replies as main posts.
Can you copy it? Our read
What you can copy is the content: long hands-on tutorials, prompts in the replies, lots of replying. What you can’t copy is the starting point: he restarted after a ban and opened with old readers. It suits people willing to get tools running themselves and then write them up at length.
- What it takes
- Keeping up with the newest Agent and coding tools, installing and running them yourself, then writing thousands of words of steps.
- Startup cost
- Time and hardware: he put his income into a 4090 and dual V100s for local deployment (self-reported), and paid ¥1,599 for a course (self-reported).
- Time to a verdict
- Watch the first two months. His new account was back to 10K in 37 days, but it brought old readers; starting from zero would be much slower (our read).
- Window
- More and more people write hands-on Agent articles, and one tool often gets several in a week; content that needs hardware, such as local deployment and multi-GPU tuning, still has few writers (our read).
Key findings
The new account never looked new
38 main posts in the first two weeks, median 6,745 views, 17 over 10,000. He says he was back to 5K in 10 days and 10K in 37. See play 2.1 (subscribers)
His biggest posts are long hands-on guides
The Codex article has 566K views and 2,439 bookmarks; among the posts we have, those over 140 characters have a median of 20,287, those of 140 or fewer 4,350. See play 2.2 (subscribers)
4.5× as many replies as main posts
1,349 main posts and 6,059 replies; 73 replies on opening day. See play 2.3 (subscribers)
Half the numbers in his recaps can’t be checked
Of 8 claims, 2 match, 1 roughly matches, 1 contradicts itself and 4 can’t be verified.
What this channel is
- Main platform
- X
- Format
- Prompt sharing + quote posts + hands-on Agent articles
- Background
- Old-school programmer, serial founder (self-reported)
- Started
- 2025-12-11 (the day the new account opened; the old one started around 2025-10 and was banned)
- Posting
- Median 4 main posts a day, 70–154 a month
- Team
- No human team seen; his bio names a company, SERVASYY LLC (self-reported)
Every number is tagged
- Fact straight from public data
- Derived calculated from facts
- Estimate from published pay ranges
- Self-reported what the creator says
- Our read our judgement
01How it happened
How the X following grew
- X 42.9K
- Fact
- Self-reported
Hollow points are follower counts he posted himself (self-reported); solid points are measured. The sign-up day and today are measured. His first account (13.9K, banned) has no exact start or ban date, so it is not drawn.
- under 1×
- 1–3×
- 3–10×
- 10× and up
Each bar is one day, in date order; its height is how many main posts (originals, quotes and articles; not replies) went out that day. This chart counts posts per day rather than listing each post. May 2–3, 2026 read zero for every account we track: a gap in the data.
Is this your account? Claim it