I spent two days tinkering with a small tool that just wouldn't work.
Last January, I tried Claude Code for the first time. I had it connected to Kimi's model, trying to build a browser extension. Every time it said "done," it wouldn't work when I tested it. All sorts of errors, endless fixing, back and forth for two days. I was about to give up—not only was I not solving it, but I was burning through tokens constantly.
Then Claude released Opus 4.6.
I bought a membership on a whim and threw the same requirements at it. Thirty minutes, from scratch, done in one go. It invoked the browser extension itself to check the results, simulated user input to test, found bugs and fixed them on its own, then told me: you can try it now.
I opened it up—and it actually worked.
In that moment, I had a strong realization: AI capabilities seemed to suddenly cross a threshold at some point. At least in certain scenarios, it wasn't just helpful anymore—it far exceeded your expectations.
Different Models, Absurdly Large Gaps
This experience happened again later.
Using Claude Code with Opus 4.6 to build a small plugin, helping a friend install a Feishu (Lark) browser automation tool. I tinkered for three or four hours; every time it said it was ready, bugs appeared when I ran it. Tried every method, just couldn't get it working.
Then ChatGPT released 5.4. I said fine, your turn. It ran for over two hours; I barely intervened in the middle, and finally it worked.
Honestly, by then I had no expectations left; I thought this might simply be impossible within current AI capabilities. But it just did it.
Same task: one model couldn't do it no matter what, switch to another and it worked. The gap between them wasn't about degree—it was about possibility versus impossibility.
Locked Out of My Computer
Then one day, something happened.
I don't know what configuration I touched while tinkering with something, but suddenly I couldn't log into my Mac. Enter username and password, hit return, wait a moment, snap—back to the login screen. Several times in a row, just couldn't get in.
In the past, I probably would have thought: ask a friend for help? Contact Apple engineers? But thinking about that process—logging into the website, queuing, scheduling, remote access, and not even guaranteed a solution—just thinking about it gave me a headache.
Then I thought: could I have the AI on another machine fix it?
The problem was, I couldn't even open my own computer; how would the AI connect?
I started searching through my network to see which device still worked. Looking around, I found one machine that had previously connected to the Zhipu AI (GLM) large model. The others either had no agent or couldn't connect either; only this one still worked. Fine, I figured I'd try it as a last resort.
Using that machine to remotely connect to the problematic Mac, I had the AI troubleshoot. As it investigated, I kept trying to log in on my end, feeding it new error messages.
Half an hour later, it found the cause. Some configuration file had issues; it helped me revert it, and I logged right in.
The feeling in that moment was very concrete. When you truly need help and the old ways can't assist you, an AI agent that can connect to your machine—even if it's not the best model—can just get things done.
Becoming More Like a Hacker
After fixing the computer, I started thinking about something else: could I use a coding agent even when not in front of the computer?
I asked Claude Code what to do. It recommended Tailscale—I had no idea what that was. After setting it up, all my machines connected to a virtual network. My phone could actually connect too.
From then on, whenever inspiration struck while walking, I could pull out my phone, connect to my computer, and have the coding agent do work for me. Actually, remotely controlling AI to do things has been possible for a while; most people just don't know how to set it up.
Later I ran into another problem: when the phone network disconnected or I exited the app, running tasks would get interrupted. Asked the AI again; it taught me to use TMUX to keep programs running in the background. After setting it up, tasks could keep running continuously, and I could check progress by connecting in anytime.
That period was quite magical. Remote control, background processes, multi-machine networking—I used to think these were programmer-only things, far removed from me. One question to the AI and it was all set up. I felt like I had really become the kind of hacker I used to imagine.
The Thing I Fear Most
But after getting used to this capability, what I feared most changed.
It's not fear of the computer breaking. It's fear of the agent going down.
Several times, the coding agent on some machine suddenly wouldn't open due to software updates or configuration changes. You've completely gotten used to going to it with problems; when it suddenly disappears, you don't even know how to fix it.
I later discovered a pattern: as long as there's at least one machine in the network with a working coding agent, I can use it to fix other machines. Just like fixing that Mac I couldn't log into before.
But what if the last agent goes down too?
Honestly, that would really make me panic.
An Idea
Thinking about this, I realized the problem wasn't just mine.
More and more friends around me are starting to use AI tools and install coding agents; when they work, it's great. But the prerequisite work of opening channels and configuring the environment is much harder than imagined. Once something breaks, most people have no idea how to fix it.
You were enjoying the convenience, then suddenly it stops working. This sense of letdown is worse than never having used it at all.
I started thinking: could we let people skip the configuration tinkering? When you need help, just open the terminal, type one command, and have an AI connect in to solve the problem for you?
This was the starting point for building Aima Service.
Aima Service
I spent some time with my team turning this idea into a product.
Running several AI agents in the backend, on standby 24/7. No need to understand technology beforehand; open the terminal, one command, and you're started.
For example, if your coding agent won't open, it helps you fix it. Want to install a new AI tool or programming environment? Hand it over. Device has some weird technical glitch? Let it investigate.
We don't guarantee 100% success. AI capabilities are still growing; some scenarios that can't be solved today might work later. But the current success rate is higher than most people think.
We've opened up substantial free quotas now; no need to pay, just log in and use it.
The Moments When AI Truly Moves People
Finally, let me share something unrelated to the product.
A friend of mine doesn't usually use AI much, and isn't clear on when to use it. Once he had a small conflict with his partner; he wanted to talk it through properly but was afraid of saying the wrong thing and making it worse. He tried asking DeepSeek. It gave some communication advice, helped him avoid several statements that could easily cause misunderstanding, and also helped him clarify what he actually wanted to express.
He later told me: AI is way more powerful than he imagined.
After that, he started actively seeking AI's help.
I've seen similar changes in myself, in him, and in many friends around me. People's views on AI often don't change because of daily convenience, but because they're moved at a moment when they truly need help and other methods aren't working.
I want more people to have moments like this. That's what Aima Service is doing.
Go to aimaserver.com and give it a try.