CAD order-taking assisted by Codex

The author describes taking CAD commissions with AI-assisted drafting and then realizing that being able to make something with AI is not the same as being able to earn money from it. Among several experiments—websites and small games in Codex, code-made animations and free tavern character cards in Claude Code, plus a recent Cursor subscription—only CAD order-taking has been commercialized.

He is explicit that he did not invent a private tool or a mysterious workflow: he used Codex together with a publicly available skill to help produce drawings. The experience showed him that AI can be practically useful, but it also made him reluctant to tell others that this is an easy way to make money. In his own order-taking, two problems stood out: clients are hard to find, and prices are hard to negotiate. He presents this as his current reality, not a verdict on the whole market.

He separates finishing a piece of work from validating demand. Playing with AI is exciting because a few rounds of conversation can turn an idea into a site, a game or an animation. Paying clients, he writes, care whether the result is usable, whether it can be delivered as requested, and whether problems can be handled—not how long he spent studying tools or prompts. Learning the tool answers how he works; winning an order still depends on why someone needs the work and why they would choose him.

A public skill lowers the learning cost, but it also makes the same method hard to defend as a reason to pick him over anyone else. If the only visible difference is that he can do it too, price becomes the comparison point, and cutting prices may not produce work worth doing over time. He does not want to reduce this to “too many people, no chance.” What he wants to answer next is more specific: which needs he understands, what communication and revision pain he can remove beyond producing a file, and whether anyone would come back after one job. He says he does not yet have mature answers, and he does not want to turn a limited experience into a how-to-get-rich tutorial.

He still values the non-commercial experiments. The attached image is a title card that reads “You can make things with AI. What then?” with the subtitle “From finishing work to finding real demand.” It does not show drawings, quotes or payment records. Order volume, unit price and total revenue are not disclosed.

Not described in detail. He says clients are hard to find, and that he spent more attention on finishing tasks than on who needs the work or how they would find him.

Volume, unit price and total revenue are undisclosed. The public skill is not named. The image is a concept card, not delivery or payment evidence. This is a different author from the existing baseline case about Codex CAD on Xiaohongshu.