Three months ago, I was just a law student who could write legal briefs and nothing else.
Three months later, I shipped a global AI-native talent marketplace.
Not because I suddenly learned to code. Because I figured out three things. And if you're using AI right now, you need to know them too.
Lesson 1: Don't Treat AI Like a Chat Box. Treat It Like a Company.
Most people use AI the same way: open ChatGPT, ask a question, get an answer, close the tab.
Next question? New chat.
This is the biggest waste.
Here's what I do instead: break AI into "departments."
My main chat is the CEO Office — strategic decisions, core conversations. Then there's Growth Engine — everything about growth. Founder Journal — content only. One more for weekly meetings, another for product and engineering.
Same idea as the "agent" feature in tools like Doubao.
This solves two problems.
First, mental chaos. Talk to AI about a dozen different topics in one chat, and good luck finding that key insight three weeks later. You can't tell which conclusion came from which context.
Second, AI compound interest. Keep one chat focused on one topic, and it gets smarter about that topic. The rules you build together, the shared understanding — the model remembers longer. It's like onboarding a new teammate: swapping every day vs. keeping the same person — completely different results.
But departments aren't enough. You need an employee handbook.
I give my AI a directory structure.
CLAUDE.md holds the core principles — things like "fix the root cause, one quality fix beats 100 guess-based patches." Architecture, operations, and SEO references each get their own file. CLAUDE.md only has the index. Read when needed.
Now the AI knows exactly what to read and when. No more loading a 500-line file on every request. Saves tokens. Saves time.
The point isn't to give AI more memory. It's to give AI structure.
Lesson 2: Don't Let Your Effort Generate Value Only Once
My biggest problem used to be: I worked hard, but the value died after one use.
Take a video. Old me thought: people watch it, that's the value.
Then I realized one piece of content can create three kinds of value.
First, it's a traffic entry point. More views = more product exposure.
Second, it's a backlink asset. Turn that video into an article for SEO. Distribute across platforms. It keeps working for you.
Third, it's a brand asset. Post to social media, then republish on your own site. Six months later, that content is still working.
Same time, same effort. Some people get one result. Others get compound interest.
This isn't just for startups.
In school — can a math concept be applied to chemistry or physics? In sales — why not document your cold outreach process and turn it into content for second-wave customer acquisition?
Later, when I built my growth system, the same pattern emerged. I started just trying to be more efficient. But halfway through I asked: could this be a product? So I designed it with features, visuals, UX in mind. Because someday it might serve more than just me.
I started training myself with one question:
What else can this produce? Can it be the foundation for the next move?
It's like chess. Not the next move. Three moves ahead.
Don't fall into the other trap though — not everything needs three values before you start. Sometimes you only see the value after you begin.
Multiplied value isn't talent. It's deliberate practice and repeated memory.
Lesson 3: Ship Your 60% Product
Finally, my turn to show off.
Before AI, I had zero coding experience. A law student whose understanding of code came from product reviews where engineers explained things to me.
In the old world, building software was completely out of reach.
But today, my 60% product is here — aiomniu.
An AI-native developer talent marketplace covering freelance, hiring, services, and community. Users can register, post projects, apply as talent. Dual review system to protect both sides. Email, Google, and GitHub login.
Everyone's welcome. Even if you just want to see what's possible.
But let me pour some cold water on you.
Don't think AI making development easier means building software is easy.
A 60% product? Yes, easier than ever. But taking it to 100%? Completely different game.
Once you have real users, you'll hit security issues. Performance issues. Architecture issues. And the "spaghetti code" AI left behind. A non-technical founder will start suffering at this point.
So here's my take:
Early stage, be confident — AI gave ordinary people the power to create.
Later stage, be humble — expertise still matters.
Honestly? I don't think I can beat Upwork. It's too strong. From day one, I wasn't trying to copy it.
I care about a different question: What kind of talent will AI-era companies need? How can individuals use AI to find new opportunities?
Nobody knows the answer yet. The biggest opportunity is right there. But so is the question: When does it really explode? Can we survive until then?
All unknowns.
But here's what I'm sure of:
People who couldn't write code used to be stuck at the idea stage. Now, AI at least gives us a shot — ship it, validate it, get your 60%.
As for the remaining 40%?
Leave it to time. And leave it to the pros.
One Last Thing
These three lessons are really one lesson.
Give AI structure. Make your effort compound. Then ship it.
A law student went from zero to a shipped product in three months. Not talent. Not luck. Just a framework for thinking.
AI lowered the barrier to action. But it didn't change the value of thinking clearly before you move.
If you're building something with AI — or still deciding whether to start — remember these three rules.
Then ship it.
Top comments (0)