← sources

How am I supposed to know what “good” or “bad” AI code looks like if I’ve never actually programmed it by hand first?

source: reddit (reddit:cscareerquestions)published 2026-10-29 02:53ingested 2026-10-30 02:37

retained text

I have 2ish years of experience as a dev. Up until now I’ve avoided using AI to write code, since I want to learn things manually. However, my org is starting to get on the AI hype train, and is urging us more and more to do “prompt engineering” and get more comfortable using AI throughout our work, and to generate more systems that align with best practices and good system design. They want us to get better at using it so we can build bigger systems quicker.

One thing that they continue to emphasize though is that we will own whatever code we generate, aka whatever code we merge into our repos. we are accountable for.

This feels like putting the cart before the horse. How am I supposed to know what “good” or “bad” system design or code looks like, if I’ve never had to struggle and implement manually first? For example, I’ve never had to build a micro-service, work with a React app, or work with SSAS. If I was asked to use an AI to generate code for these things tomorrow, I would have absolutely no idea how to vet the output. I.e: Whether it generated is good or not, how scalable it is, what edge cases it’s accounting for (or not), how efficient the approach is, etc. I have no relevant prior experience to fall back on.

No idea how to approach this at work now. I don’t feel comfortable generated vast amounts of code that I don’t fully understand. But I feel like our AI usage is going to be more closely monitored going forward.

document fetched 2026-10-30 00:23 · canonical https://reddit.com/r/cscareerquestions/comments/1xi8jr2/how_am_i_supposed_to_know_what_good_or_bad_ai