My vibecoding two day session according to Claude>Overall>Total Lines: 9,997 lines of TypeScript>Total Files: 55 TypeScript files> Breakdown>Source Code (src/)>Lines: 3,827 lines>Files: 27 files>Purpose: Production application code>Test-to-Code Ratio>For every line of source code, there are 1.52 lines of test code>This indicates excellent test coverage!>Key Insights>The project has nearly 10,000 lines of TypeScript code>Test code is ~60% larger than source code, showing strong commitment to quality>With 238 passing tests and 3 skipped, you have excellent test coverage>The largest files are test files, which is a good sign of comprehensive testing
You're better than clankers. Use your brain to code.
>using Typeshit
omg special weeku
>>106999935did you actually check that the talking GPUs weren't lying though?
>>107000921normal week
>>107000921Cursed image
>>107000921You can use Codex + Claude to examine code and both complement each other. For this project, I was using ChatGPT web, it was handholding me and guiding me a lot, meanwhile Claude was doing the grunt work of reading and writing files. If there was a fuck up I would made them talk to each other and they eventually reach a fix.I've been testing the endpoints myself and so on, Claude is insane
>For every line of source code, there are 1.52 lines of test codeDo you really need this? It sounds retarted. It seems like an overkill, bloated even. Even If your code receive a lot of data from external sources.
>>106999935haha OP I love horse girl XD
>>107002102I even set up database tests with docker composeI fucked up big time at the beginning by not using an ORM, that caused major issues, syntax mismatch and other things.Look, from what I've learned is that if you want to vibecode and be on full autopilot, feature developer, you also need to provide the safest environment possible to catch bugs and errors so the AI has a clue as well
