I finished an English series on the way I think ordinary people can start using AI for real work.
The point is not to become an AI expert first.
The point is to have one place where you can say what you want, give the tool access to the right folder, and check the result.
Anything important still needs a human pause: publishing, deleting, paying, or authorizing.
My preferred starting point is simple:
One project folder. One command-line window. One plain-English request. One habit of checking before confirming.
That is the "AI workbench" idea.
Why I started with the command line
Most people hear "command line" and think it means programming.
I think that is the wrong frame.
For an AI tool like Codex, the command line can be the cleanest interface because it does not start with a pile of buttons.
You type the goal:
I want to organize the files in this folder and turn the useful material into a public article.
Please inspect the folder first.
Tell me what looks public, what looks private, and what you would do next.
Do not move or publish anything until I approve the plan.
This is not a developer-only request. It is a normal human request with a goal, boundary, and confirmation rule.
What the series covers
The series is for people who can use a computer or phone, but do not want to memorize AI jargon before doing anything useful.
It covers:
- Your Personal AI Workbench
- Why I Think One Command-Line Window Is All You Need
- AI Tools Actually Come in Four Types
- Web, App, Client, IDE, CLI — What Is the Difference
- Why Ordinary People Can Use CLI Too
- The Minimum AI Glossary
- Foreign Models, Accounts, Networks, and Real-World Limits
- Ten Things Ordinary People Can Do with AI Right Away
- When Not to Use AI
- My Minimum AI Workbench
- AI Glossary for Ordinary People
The practical test
My test for an AI tool is not whether it has the most impressive demo.
It is whether it can reach the work.
The questions I ask are practical:
- Can it read the current folder?
- Can it edit the file I am actually working on?
- Can it run a check?
- Can it open or name the next page I need?
- Can it stop at the moment where I need to confirm an account, payment, deletion, authorization, or public publish?
If a tool cannot reach the work, it may still be useful for conversation. But it is not a workbench yet.
The safety boundary
This series is not an argument for handing judgment to AI.
The more useful an AI tool becomes, the more important the boundary becomes.
I do not want AI to decide medical, legal, investment, payment, publishing, or deletion questions for me.
What I want is preparation, inspection, comparison, drafting, organization, and explanation — then a stop where a human decision is needed.
That habit matters more than any specific product name.
A first prompt
If you already have Codex installed, start in an empty folder and try this:
I want to start using AI to help me get things done, but I am not a programmer
and I do not want to learn a bunch of jargon first.
Please tell me: in this folder, what can you help me with?
Can I ask you to read files, change content, make a plan, explain something, or check for errors?
Give me one small first task I can do here — steps small enough that you can guide me through each one.
The question is not "how much AI theory do I know?"
The better question is: can I describe a real task clearly enough that the tool can help me move one step?
Full series:
https://marlinbian-site.pages.dev/en/tutorials/ai-workbench-for-everyone/
Originally published on my personal site:
https://marlinbian-site.pages.dev/en/tutorials/ai-workbench-for-everyone/
More links: GitHub · YouTube · LinkedIn · Bluesky · Mastodon · Discord