Last Wednesday, Ah Lin asked one question that stopped me cold

Last Wednesday afternoon, I was helping a friend who runs a home-baking side hustle polish her copy. Ah Lin suddenly looked up and asked: "Are the AI tools you use actually reliable? What if they just cut you off one day?" I froze for 3 seconds—I'd only been thinking about how to get AI to write better, never about what company was behind it, or what it was up to.

Cal Newport's call: investigate the AI labs

Cal Newport is a computer science professor at Georgetown—the Deep Work guy. He just put out a piece with one core message: the big companies building AI tools (OpenAI, Anthropic, Google DeepMind, etc.) are stacking up more and more power, and we should seriously start scrutinizing them.

He's not saying don't use AI tools. He's saying: how these companies train models, set prices, treat users—a lot of it is opaque. We're the ones stacking our businesses on these tools, so we're not bystanders—we're stakeholders.

Concrete scene: my friend Xiaolin started using a domestic AI to write Xiaohongshu (aka RedNote) copy last November. In two months she grew 10,000 followers. End of February, that company suddenly changed its pricing—she did the math and her monthly cost jumped from 30 yuan to 800 yuan. She couldn't even find a customer service email.

What you can do today: the 5-minute version

Money: 0 yuan
Time: 15 minutes (skim news + fill in a checklist)
Technical barrier: Just browse a webpage, no jargon
First step: Open any AI tool you've paid for or use daily, click "About Us," and write down three questions—where's this company headquartered? Who's funding it? When was its last funding round?

If you're... I'd...

Just starting out, still on free tiers: You don't need to panic yet. But I'd suggest at least writing down one name—your main tool—so if it suddenly dies, you can find a replacement within 24 hours instead of freezing up.

Already have 1-2 paying clients, starting to use AI for real work: I'd suggest spending an hour building a "vendor map"—categorize the AI tools you depend on as core / supporting / nice-to-have, and keep at least one backup for each tier. I skipped this step once, and when a tool suddenly throttled me, I pulled a 36-hour sprint with a contractor to recover.

Scaling up, spending over 3,000 yuan/month on AI: I'd suggest reading Cal Newport's original piece (English, 8 minutes), then spending half a day tracking news about your main vendors. This investment is worth as much as what you put into ads or product selection—when they shift, you shift with them.

One last honest thing: I'm not trying to make you distrust AI tools. Quite the opposite—I use them every day. But I made a mistake before—I thought a tool was just a tool, you pay and use it. The truth is, every AI tool stands on a company, and that company has its own fate, its own troubles, its own variables. Before you hand your business to one, at least get to know it.