Hacker News new | ask | show | jobs
by paulhebert 134 days ago
I think there are a number of elements:

- What you are working on. AI is better at solving already solved problems with lots of examples.

- How fast/skilled you were before. If you were slow before then you got a bigger speed up. If AI can solve problems you can’t you unlock new abilities

- How much quality is prioritized. You can write quality, bug free code with AI but it takes longer and you get less of a boost.

- How much time you spend coding. If a lot of your job is design/architecture/planning/research then speeding up code generation matters less

- How much you like coding. If you like coding then using AI is less fun. If you didn’t like coding then you get to skip a chore

- How much you care about deeply understanding systems

- How much you care about externalities: power usage, data theft, job loss, etc.

- How much boilerplate you were writing before

I’m sure that’s not a complete list but they are a few things I’ve seen as dividers
2 comments
ggregoire 134 days ago
> How much you like coding. If you like coding then using AI is less fun.

I'm surprised this is never brought up here on "Hacker" News. I've been reading HN for 14 years and all this time I thought people here enjoyed programming. Turns out the majority hates it apparently.

link
paulhebert 134 days ago
Yeah, it’s strange to me. One of the reasons I chose this career is because I enjoy programming.

I don’t get the same satisfaction when using AI coding tools. I feel like I’m the manager of a faceless team of mediocre employees

link
paulhebert 134 days ago
A few more:

- How much do you prioritize speed?

- Do you have a big backlog of dev tasks ready to go?

- What are the risks if your software doesn’t work?

- Are you working on a green field or legacy project? Prototypes or MVPs?

link
paulhebert 134 days ago
- Do you prefer working as a manager or an individual contributor? Are you used to owning the code or managing others who write codd?

- How does your brain work? Do you rely on flow states? Do you get distracted waiting for the LLM response? How do you handle context switching?

link