Scene hook
Last week, while I was revising client copy in a coffee shop, a thought hit me: what happens to my work if the AI I rely on suddenly gets more expensive, rate-limits me, or locks up my history? I’ve gotten stuck here too. For a while, I had drafts, prompts, and client info piled into a single platform.
What this is + who’s already using it
After reading Mozilla’s report on the state of open AI, what stayed with me wasn’t really “the tech is getting stronger.” It felt more like a reminder for people like us: don’t just look at which tool seems smartest. Also look at whether it’s portable, whether we can keep our own data, and whether we can switch vendors later. Not everyone needs to jump into open models right away, but it helps to know there’s no longer just one path. At 8 p.m. on a Tuesday last month, in a coworking office in Binjiang, Hangzhou, Lin Ran, who sells educational content online, told me she now keeps her course outlines and client Q&A in separate places: public-facing content goes through mainstream AI platforms, while sensitive material stays in tools she can export from herself. It’s not a flashy move, but it’s steady.
What it costs to copy today
If I were worried about getting boxed in by a single platform, I’d start with a super-light version: RMB 0-99, 30 minutes, and the technical barrier is just being able to register accounts for two tools and save documents. No coding needed. The first step isn’t studying model parameters. It’s opening the “Export,” “Download history,” or “New document” button in the tool I already use, then backing up my common prompts, client FAQs, and sales messaging first. If I don’t try more today, that’s still fine. I just wouldn’t keep everything in only one place.
Advice by stage: just starting / 1-2 clients / scaling up
If I were just starting, I’d focus first on whether I can export my content, not on chasing the newest model.
If I already had 1-2 clients, I’d store client information, pricing language, and delivery templates separately so platform changes don’t throw me into panic.
If I were scaling up, I’d start looking at whether my setup supports swapping out the underlying model, because at that point I probably wouldn’t be missing a smarter AI so much as a system that doesn’t lock my whole business in.