Wang Yuquan's Deep Dive into "Raising Shrimp": After Two Weeks of Staying Up Late, OpenClaw Changed My Research Workflow
Nizwo OpenClaw Column shares first-hand OpenClaw usage experiences. Follow us to master the Lobster.
"This is the last time I urge them to pay attention — a silent 'isolation' is beginning."
In May 2026, Wang Yuquan (founding partner of Haiyin Capital) wrote a long commentary article.
The title is direct: "Wang Yuquan Actually Stayed Up Late Raising Lobsters and Horses! It Turns Out a Silent 'Isolation' Is Beginning."
The core of the article is not reviewing OpenClaw or Hermes technical details, but observing a social phenomenon:
"Standing at the forefront of the AI era, the behaviors of two groups of people are completely different."
What did Wang Yuquan do?
According to the article description, Wang Yuquan stayed up late for two weeks, deeply using OpenClaw (nicknamed "raising shrimp") and Hermes Agent (nicknamed "raising horses").
What exactly did he do?
- Automatically collecting materials
- Analyzing data
- Modifying code
- Handling complex tasks
"Frequently staying up until 2 or 3 in the morning."
Note: This is a senior investor and former university professor personally staying up late to use AI Agent to get work done, not the level of "experiencing it once and writing a review."
The divergence of two groups
Wang Yuquan described a forum scenario in the article:
"At a previous forum, I almost got into an argument with another expert."
What did they argue about? The other person denied AI, with reasons: 1. OpenClaw is immature 2. OpenClaw is unreliable 3. Some even said OpenClaw is another form of virus
Wang Yuquan's observation:
| Group 1 (Skeptics) | Group 2 (Practitioners) |
|---|---|
| Focus on AI's problems | Focus on AI's capability breakthroughs |
| Have installed OpenClaw, know it's immature | Don't care if it's mature; problems can be solved |
| Use "immaturity" as a reason not to act | Only care about "what else can be done with it" |
| AI is just an object of criticism | AI is a tool used every day |
| Stuck in the debate of "will AI replace people" | Already using AI to replace their own repetitive labor |
Wang Yuquan's conclusion:
"Seemingly just a difference in attitude, but it will lead to two completely different results."
How did OpenClaw change the research workflow?
Although Wang Yuquan didn't detail his workflow in the article, based on the description and OpenClaw's capabilities, we can infer:
Traditional research workflow (Wang Yuquan's previous method)
1. Manualy search for materials (Google/academic databases/news)
↓ (Time consumption: hours to days)
2. Manualy read, filter, take notes
↓ (Time consumption: days to weeks)
3. Manualy analyze, write first draft
↓ (Time consumption: days)
4. Manualy revise, polish
↓ (Time consumption: days)
5. Publish
Total time: 1-2 weeks (one topic from research to publication)
Workflow after using OpenClaw
1. Agent automatically collects materials (multi-threaded, covering Chinese and English)
↓ (Time consumption: minutes to tens of minutes)
2. Agent automatically reads, extracts key information, categorizes
↓ (Time consumption: minutes)
3. You review key information and direction
↓ (Time consumption: tens of minutes)
4. Agent writes first draft based on your feedback
↓ (Time consumption: minutes)
5. You revise, polish
↓ (Time consumption: hours)
6. Publish
Total time: 1-2 days (efficiency improved 5-10×)
Core change: Not "AI replaces human writing articles," but "AI does the most time-consuming material collection and initial screening work, while humans focus on judgment and decision-making."
Why OpenClaw?
The reason Wang Yuquan chose OpenClaw instead of other AI tools (ChatGPT, Claude, etc.) is:
1. 7×24 operation, no need for human supervision
OpenClaw runs on local computer or server, it continues working while you sleep.
Traditional AI assistant (ChatGPT web version): You ask one question, it answers one question; the session ends when you close the tab.
OpenClaw: You give it a goal, it autonomously decomposes tasks, autonomously executes, autonomously verifies, you wake up and see the results.
2. Can operate your computer
OpenClaw integrates: - Browser control (automatic search, automatic form filling, automatic clicking) - Terminal command execution (automatic script running, automatic file processing) - File system read/write (automatic material organization, automatic report generation)
This means it can realy "help you get work done," not just "answer your questions."
3. Local operation, data doesn't leave home
As an investor, Wang Yuquan researches undisclosed projects and industry internal information; data privacy is very important.
OpenClaw can run completely localy (using local large models, not calling cloud APIs), ensuring data doesn't leave local.
"Raising Shrimp" vs "Raising Horses": Wang Yuquan's choice
The article title mentions both "raising shrimp" (OpenClaw) and "raising horses" (Hermes Agent).
According to search results, Wang Yuquan is using both Agents simultaneously, but with different focus:
| Comparison | OpenClaw (Raising Shrimp) | Hermes Agent (Raising Horses) |
|---|---|---|
| Positioning | Swiss Army Knife, wide functionality, wide integration | Self-evolving apprentice, gets smarter the more you use it |
| Suitable scenarios | Quick start, wide integration | Long-term companionship, deep customization |
| Wang Yuquan's usage | Research workflow automation (material collection, first draft generation) | Deep analysis, complex reasoning |
| Onboarding difficulty | Low (install and use) | Medium (requires configuration and training) |
Wang Yuquan's strategy: Use both, select different Agents for different tasks.
The great divide: Two kinds of people, two kinds of results
The core argument of Wang Yuquan's article is not "OpenClaw is very useful," but:
"A silent 'isolation' is beginning."
This "isolation" refers to:
People who use AI and people who don't use AI will produce a huge capability gap in the next 5-10 years.
It's not a question of "will AI replace people," but:
- People who use AI: 1 person + AI = output of 10 people in the past
- People who don't use AI: 1 person = output of 1 person in the past
Over time, the gap will expand exponentially.
Wang Yuquan said in the article:
"Companies will calculate and realize that for the same work, AI costs less and costs will continue to decline, and AI's scalability is stronger. Then companies will naturally transition to Agent-based operations. If employees' work methods are incompatible with Agent transformation at this time, they will likely be laid off."
Relationship with Nizwo
Wang Yuquan uses his own computer to run OpenClaw, needing the computer to be on all the time.
Nizwo's value: Gives you a hardware device dedicated to running OpenClaw, plug in and use, 7×24 online, no need to keep your main computer on all the time.
Wang Yuquan's workflow (optimized with Nizwo):
Daytime: Use OpenClaw on Nizwo to automatically collect materials, generate first drafts
↓
Evening: Review, revise, polish
↓
Sleep: OpenClaw continues running on Nizwo (7×24)
↓
Next day: See results, continue next round
Nizwo = Dedicated hardware base for OpenClaw.
Something is happening
Wang Yuquan's article, together with Google Gemini Spark release and Alibaba Qoder 1.0 release (same week), points to the same trend:
In 2026, AI Agents move from the "demo stage" to the "practical stage."
- Before: Everyone watched AI demo videos, thought "wow so amazing," but couldn't use it themselves
- Now: Professionals like Wang Yuquan are realy using AI Agents to get work done, and efficiency improved 5-10×
The significance of this turning point may be seriously underestimated.
Nizwo OpenClaw Column shares first-hand OpenClaw usage experiences. Follow us to master the Lobster.
/uploads/images/829af88160544e0385969a772a29700b.webp