Things that helped me get out of the AI 10x engineer imposter syndrome • Hacker News Board

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#21

Post by mrbungie » Tue, Aug 05, 2025, 2:29 PM UTC

jf22 wrote: post #2

> What LLMs produce is often broken, hallucinated, or below codebase standards. With enough rules and good prompting this is not true. The code I generate is usually better than what I'd do by hand. The reason the code is better all the extra polish and gold plating is essentially free. Everything I generate comes out commented great error handling, logging, SOLID, and united tested using established patterns in the…

> The code I generate is usually better than what I'd do by hand.

I'm always baffled by this. If you can't do it that well by hand, how can you discriminate its quality so confidently?

I get there is a artist/art consumer analogy to be made (i.e. you can see a piece is good without knowing how to paint), but I'm not convinced it is transferrable to code.

Also, not really my experience when dealing with IaC or (complex) data related code.

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#22

Post by micromacrofoot » Tue, Aug 05, 2025, 2:29 PM UTC

jf22 wrote: post #2

> What LLMs produce is often broken, hallucinated, or below codebase standards. With enough rules and good prompting this is not true. The code I generate is usually better than what I'd do by hand. The reason the code is better all the extra polish and gold plating is essentially free. Everything I generate comes out commented great error handling, logging, SOLID, and united tested using established patterns in the…

I've been finding actual human-written bugs and correcting them with Claude, so I find the "often broken" claims a load of nonsense... I've been fixing dozens of minor bugs in our codebase that no one's been arsed to fix for years due to bigger priorities (which tbh is generating more features and tech debt).

It may change in the future, but AI is without a doubt improving our codebase right now. Maybe not 10X but it can easily 2X as long as you actually understand your codebase enough to explain it in writing.

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#23

Post by echelon » Tue, Aug 05, 2025, 2:29 PM UTC

the__alchemist wrote: post #7

jf22 wrote: post #2

> What LLMs produce is often broken, hallucinated, or below codebase standards. With enough rules and good prompting this is not true. The code I generate is usually better than what I'd do by hand. The reason the code is better all the extra polish and gold plating is essentially free. Everything I generate comes out commented great error handling, logging, SOLID, and united tested using established patterns in the…

I am pattern matching your last statement with what I've seen with my teammates who are more AI-oriented: I suspect this is a matter of making the metrics the goal. I would rather maintain something that is simple, works, and have targeted comments than something messy that meets the metrics you list.

I don't get all the prompt vibe coding going around. I don't use prompts to generate code.

I use "tab-tab" auto complete to speed through refactorings and adding new fields / plumbing.

It's easily a 3x productivity gain. On a good day it might be 10x.

It gets me through boring tedium. It gets strings and method names right for languages that aren't statically typed. For languages that are statically typed, it's still better than the best IDE AST understanding.

It won't replace the design and engineering work I do to scope out active-active systems of record, but it'll help me when time comes to build.

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#24

Post by exe34 » Tue, Aug 05, 2025, 2:30 PM UTC

drgiggles wrote: post #4

jf22 wrote: post #2

> What LLMs produce is often broken, hallucinated, or below codebase standards. With enough rules and good prompting this is not true. The code I generate is usually better than what I'd do by hand. The reason the code is better all the extra polish and gold plating is essentially free. Everything I generate comes out commented great error handling, logging, SOLID, and united tested using established patterns in the…

Easily 99% of comments generated by LLMs are useless.

They comment on the how, not the why.

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#25

Post by intended » Tue, Aug 05, 2025, 2:30 PM UTC

jf22 wrote: post #2

> What LLMs produce is often broken, hallucinated, or below codebase standards. With enough rules and good prompting this is not true. The code I generate is usually better than what I'd do by hand. The reason the code is better all the extra polish and gold plating is essentially free. Everything I generate comes out commented great error handling, logging, SOLID, and united tested using established patterns in the…

Could you share an example ?

These conversations on AI code good, vs AI code bad constantly keep cropping up.

I feel we need to build a cultural norm to share examples places of succeeded, and failures, so that we can get to some sort of comparison and categorization.

The sharing also has to be made non-contentious, so that we get a multitude of examples. Otherwise we’d get nerd-sniped into arguing the specifics of a single case.

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#26

Post by shortrounddev2 » Tue, Aug 05, 2025, 2:30 PM UTC

drgiggles wrote: post #4

jf22 wrote: post #2

> What LLMs produce is often broken, hallucinated, or below codebase standards. With enough rules and good prompting this is not true. The code I generate is usually better than what I'd do by hand. The reason the code is better all the extra polish and gold plating is essentially free. Everything I generate comes out commented great error handling, logging, SOLID, and united tested using established patterns in the…

Easily 99% of comments generated by LLMs are useless.

I think its because LLMs are often trained with data from code tutorial sites and forums like stackoverflow, and not always production code

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#27

Post by mrugge » Tue, Aug 05, 2025, 2:30 PM UTC

micromacrofoot wrote: post #9

drgiggles wrote: post #4

Earlier quoted context omitted.

Easily 99% of comments generated by LLMs are useless.

They're often repetitive if you're reading the code, but they're useful context that feeds back into the LLM. Often once the code is clear enough I'll delete them before pushing to production.

do you have proof of this being useful for llm? wouldn't you rather it re-read the actual code it generated instead of assuming that the potentially wishful thinking or stale comment is going to lead it astray?

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#28

Post by symfrog » Tue, Aug 05, 2025, 2:31 PM UTC

rob_c wrote: post #16

symfrog wrote: post #6

Earlier quoted context omitted.

Do you have a link to some of the code that you have produced using this approach? I am yet to see a public or private repo with non-trivial generated code that is not fundamentally flawed.

That's disingenuous or naive. Almost nobody decides to expressly highlight the section of code (or whole files generated by ai) they just get on with the job when there's real deadlines and it's not about coding for the sake of the art form...

I am talking about correctness, not style, coding isn't just about being able to show activity (code produced), but rather producing a system that is correctly performing the intended task

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#29

Post by r3trohack3r » Tue, Aug 05, 2025, 2:31 PM UTC

jf22 wrote: post #2

> What LLMs produce is often broken, hallucinated, or below codebase standards. With enough rules and good prompting this is not true. The code I generate is usually better than what I'd do by hand. The reason the code is better all the extra polish and gold plating is essentially free. Everything I generate comes out commented great error handling, logging, SOLID, and united tested using established patterns in the…

I do think a lot of the discourse in this space can be summed up as: people are arguing about two non-overlapping segments of a distribution having no idea the other segment even exists; instead they just assume the other side is [hype/pessimistic].

Re: Things that helped me get out of the AI 10x engineer imposter syndrome

#30

Post by tptacek » Tue, Aug 05, 2025, 2:31 PM UTC

mrbungie wrote: post #21

jf22 wrote: post #2

> What LLMs produce is often broken, hallucinated, or below codebase standards. With enough rules and good prompting this is not true. The code I generate is usually better than what I'd do by hand. The reason the code is better all the extra polish and gold plating is essentially free. Everything I generate comes out commented great error handling, logging, SOLID, and united tested using established patterns in the…

> The code I generate is usually better than what I'd do by hand. I'm always baffled by this. If you can't do it that well by hand, how can you discriminate its quality so confidently? I get there is a artist/art consumer analogy to be made (i.e. you can see a piece is good without knowing how to paint), but I'm not convinced it is transferrable to code. Also, not really my experience when dealing with IaC or (comple…

What an odd question. For the exact same reason people who write prose professionally usually have someone else edit their work: because editing your own work is harder, and everybody slips up sometimes.