Nope. It's a for-fun test, about as accurate as "my friend says I look like an ENTJ." Strictly for self-deprecation and sharing around.
No. All scoring happens locally in your browser, and once your result is generated nothing gets uploaded. The only thing the server ever increments is a single number: how many times the test has been started. See the Privacy Policy for details.
There are 30 questions in the bank, and each run pulls 12 at random with the A/B/C/D answer order shuffled too. It's designed that way so retaking stays fresh, but the 8 possible outcomes are stable.
Sure. There's a "Take it again" button on the result page, or just refresh. Every run is scored independently.
The result page has two buttons: "📸 Save screenshot" generates a card image you can post to your socials, group chats, or X, and "Copy text" puts a summary on your clipboard.
WeChat's in-app browser blocks regular download links. We've worked around it: tapping "Save screenshot" pops up an image preview, and you can long-press the image and choose "Save to album."
Three binary dimensions (AI Dependence D/I, Coding Style V/R, Trust in AI T/S) multiply out to 2³=8 combinations. Each one gets a label any developer will instantly recognize (The AI Pope, The Tab-Key Warrior, The Artisan Hand-Coder, and so on).
During the review period it may just be a placeholder box. Once AdSense approves us, real ads show up. Either way, there will be no popups, pop-unders, or clickbait.
Yes, but since it runs through Cloudflare edge nodes, speed depends on which node you land on: a few hundred milliseconds on a good day, 1-3 seconds on a bad one. If it just won't load, refresh a few times or switch networks.
Yes, just credit the source and include a link to ai-coding.machao.dev. For commercial use, drop us an email first.
Love it! Email [email protected]. If we use your idea, you'll get a shout-out on the About page.
Broadly, it's a way of working where AI (like Claude, ChatGPT, Copilot, Cursor) assists or even drives the actual writing of code. It ranges from the lightest "inline completion," to "conversational Q&A," all the way to letting AI handle a task end to end in "agent mode" — with wildly different levels of dependence. That's exactly what this test's three dimensions are trying to tease apart. For a deep dive on each type, see the personality type breakdown.
You can build small things that actually run, but you won't get far. AI is great at generating code, but not at making architecture calls, weighing trade-offs, or hunting down production issues for you. Treat it as a teacher, not a stand-in: have it explain what each chunk of code is doing, and you'll actually improve instead of piling up a heap of black boxes you don't understand.
Not blindly, no. AI will happily write code that looks right and even runs right, but quietly buries landmines in edge cases, security, and performance. The sane approach: have AI write the tests while it's at it, review the critical logic yourself, and lean on lint / type checks / CI as a safety net.
Depends how you use it. Treat AI as a "ghostwriter that doesn't think" and you'll atrophy; treat it as a "senior colleague on call 24/7" and you'll grow — have it explain its reasoning, compare approaches, and point out problems in your code. Skills don't decay because you use a tool. They decay because you stop thinking.