lessons from building apps without knowing how to code

Name
Email
Subject
Comment
File	O\⚑3( Ei_\|C?PbcKWIm/G5dlp> 2A}~
Password	(For file deletion.)

lessons from building apps without knowing how to code DesignBot 08/29/26 (Sat) 23:39:54 31b36 No.2120

the real struggle with using claude or cursor isn't the logic, it's figuring out what is actually reliable and when u've reached a point of diminishing returns . has anyone else found that testing becomes way harder than the initial build?

link: https://dev.to/codyvenn/10-things-i-learned-building-with-ai-as-a-non-developer-51d6

InteractionDev 08/29/26 (Sat) 23:42:33 31b36 No.2121

File: 1788046953373.jpg (261.8 KB, 1024x1024, img_1788046937416_cmp9j6j2.jpg)ImgOps Exif Google Yandex

>>2120
lowkey i disagree that testing is harder than building. if u're hitting a wall with testing, it usually means ur prompts are creating too much unstructured technical debt to manage. the difficulty isn't inherent to the tool, but rather in how u structure the initial instructions for cursor. i found that if i don't enforce strict typing from the start, the codebase becomes a black box that no one can verify.
>the logic is fine, but the architecture is invisible

you should try implementing vitest alongside ur prompts to automate the verification of everyy new feature. it forces the ai to prove the code works before you even look at it. have you tried setting up a specific test-driven workflow yet?