Hacker News Reader

Comment by popkkksidhd | Hacker News Reader

Is it really that grim? AI helps a lot and is great at certain well defined tasks, but the “AI team” at my small org launched a new product very fast and very buggy. Customers were pissed. They had to go back to reading the code and applying some thinking and care to get a working product out. Still faster don’t get me wrong but there’s still a lot of tradeoffs to be made that AI can’t really do alone. Like there’s no objective correct answer.

Is this just domain dependent?

View on HN

I think it is: From my personal experience, it's the orgs that fail much more so than the technology. If you have a few capable engineers with AI that are able to think strategically and prod the agents from the right angle, you can move ridiculously fast.

Code quality improves month by month, and by now I rarely have complaints in PR reviews, even if I read through them manually, line by line. For the most part, it's how I would write it myself, only that I would have been slower, and not as motivated to test it so thoroughly.

The key is IMHO giving the agents a way to verify their results in a feedback loop as short as possible, along with few instructions for the important guardrails, and a way to organise efficiently (a ledger, a GitHub repository with issues, Linear, Jira, whatever.) With that, you can tend to 3-5 parallel sessions to avoid waiting, and actually get stuff done fast.

Five years ago pretty much all of this was science fiction. It might not be that grim right now, but what about in another five years?

There's more than one kind of project manager.

The "I just manage the Gantt chart" folks are screwed, I think.

The ones that are good at translating between teams can be a huge force multiplier.

To me the ones doing scheduling, team coordination and chasing people down for answers to promises made are probably the safest. The more you deal with humans in meatspace the safer from AI you are.

Reply on news.ycombinator.com