Wayne Duan

How new technology becomes real business in messy markets.
Field notes from AI, care, health, China, and offline industries.
What makes a toy or a business.

[Wayne Duan @GlobalAgingAI]

I work in one of the most traditional, offline, human-heavy industries: care.

After thousands of trials, I’ve learned that the hard question is not:
“Can AI do this?”

The hard question is:
“Can AI create value inside a real workflow, with a real buyer, under real constraints?”

Most AI products are toys because they stop at the demo.

Real AI businesses survive:
- messy workflows
- low-trust environments
- non-technical users
- unclear buyers
- fragmented payments
- regulation
- human resistance

I write about AI deployment in care, health, and messy human systems.

If you are building, investing in, or thinking about AI, let’s discuss the question that matters:

Which AI is just a toy?
And which AI can actually make money?

[Wayne Duan @GlobalAgingAI]

刚开始用Claude Dynamic Workflows，很容易忽视以下3个细节：

1）rubric 写不清楚

"帮我筛出最重要的 bug"，这句话对工作流来说毫无意义。每个涉及评判、筛选、排序的工作流，都需要你明确告诉它标准是什么，比如："优先级依据：用户受影响范围 > 是否阻塞核心流程 > 是否有临时方案"。
rubric 越具体，输出质量越高。这一步偷懒，后面全白费。

2）不设 token 预算就全量跑

动态工作流会启动大量子 agent，每个都要加载上下文，token 消耗可以是普通任务的数十倍。正确做法：先用10%的数据量试跑一次，确认方向和成本。再在 prompt 里加上 `"token 预算控制在 X 以内"`，全量再跑。

3）好用的工作流跑完就扔，是最大的浪费

Claude 生成的工作流是可以保存的。跑完后按 `s`，存入 `~/.claude/workflows`。下次同类任务直接调用，不用重新描述，不用重新调参。
好的工作流会随着使用次数不断优化，这是复利，不是一次性消耗品。

[Wayne Duan @GlobalAgingAI]

@lvman77 支持，点赞👍

[Wayne Duan @GlobalAgingAI]

@JapanTank 还差一次成功的入侵

[Wayne Duan @GlobalAgingAI]

@better_christal 这个赛道可以，主打反差感

[Wayne Duan @GlobalAgingAI]

放纵一下，吃个甜的！

[Wayne Duan @GlobalAgingAI]

A famous European geriatrician asked me today:

“How do I turn 30 years of research into a real business without exhausting myself?”

That question hit me harder than I expected.

He is already retired, but not really retired.

After decades of research, he started a company to commercialize one of his major outcomes.

But at the end of our meeting, he said something very honest:

“The company is basically just me as the only full-time person, plus a few part-time advisors. I fly around the world every year. I’m tired. And honestly, I’m not making much money.”

I laughed and told him:

“One-person company is a very fashionable model now. It is also my goal. But there is one condition: you must know how to use AI.”

Then I suddenly remembered my first startup 10 years ago.

We were trying to build a copilot for clinicians, helping doctors commercialize their patents and research outcomes.

But back then, everything was heavy.

An MVP took at least 6 months.

Initial funding started from around $50,000.

You needed product people, designers, engineers, sales, and a lot of patience before even knowing whether the idea worked.

Today, the game is completely different.

With AI, one expert can turn decades of knowledge into a working MVP in a week.

The startup cost can be close to zero.

The first version no longer requires a full team.

It requires clarity, domain expertise, and the ability to work with AI.

This is a massive shift for healthcare, elderly care, and every knowledge-heavy industry.

In the past, experts needed a team before they could build a product.

Now, experts can build the product first, then decide whether they really need a team.

The bottleneck is no longer building.

The bottleneck is whether the expert can think like a product person.

[Wayne Duan @GlobalAgingAI]

@MOBILEiwasi 对嘛，我们也讨厌政府。

[Wayne Duan @GlobalAgingAI]

@e5ivxh 租房不也得花钱吗？租金付出去，房屋资产却是房东的。租房是给房东打工，买房给银行打工。都差不多。😄

[Wayne Duan @GlobalAgingAI]

@hFqdFKdBQrG9htI 亚洲小朋友刚出生时，看着都很像。很乖的宝宝

[Wayne Duan @GlobalAgingAI]

@Gingeramyashley 赢家思维 👍

[Wayne Duan @GlobalAgingAI]

@on_enten 韩国或者东南亚能够替代中国的食物供应吗？

[Wayne Duan @GlobalAgingAI]

@AI_Jasonyu 联系国外朋友代充

[Wayne Duan @GlobalAgingAI]

哪些朋友能看到这条有趣的post😄

[Wayne Duan @GlobalAgingAI]

@macha3142 说实话，看起来有些普通

[Wayne Duan @GlobalAgingAI]

在中国，感觉身边到处都是失意的生意人和投资客。

[Wayne Duan @GlobalAgingAI]

Most roads don't come with certainty.

You move forward.
You adjust.
You learn.
And one day you realize:
The road was shaping you more than leading you.

[Wayne Duan @GlobalAgingAI]

@Nicotama222 AIに聞いたら、 「全員でいいねし合えば万バズの確率は上がります」 と言われました😆
まずは実験してみましょう！

[Wayne Duan @GlobalAgingAI]

@lianjinshushu_ 握手！互关，互动 😄

[Wayne Duan @GlobalAgingAI]

Hot take! How fast you walk might predict when you die.

From a study of 150,000 people across 25 countries, tracked for 24 years.