26
1
The Problem With LLMs (www.deobald.ca)
27
1
28
3
Cheap design (dottedmag.net)
29
4
30
1
31
2
32
2
Magic Words (daverupert.com)
33
4
A Language For Agents (lucumr.pocoo.org)
34
2
35
6
36
2
37
2

This is the paradox: AI reduces the cost of production but increases the cost of coordination, review, and decision-making. And those costs fall entirely on the human.

38
1
39
1
40
1
41
1
Claude Code Swarms (addyosmani.com)
42
1
My AI Adoption Journey (mitchellh.com)
43
2
44
1
45
2
46
1
47
1
Agentic Engineering (addyosmani.com)
48
1
49
1

TL;DR

So, what's the verdict? Can AI replace an experienced developer for debugging?

No, of course not. AI is very good at pattern recognition, and it can be very, very useful. It will do exceptionally well on standard stuff: schema validation mishaps, forgotten null checks, common runtime errors, you've seen it for yourself already. Personally, my debugging almost always starts with the LLM these days. With literally the prompts above: "here's what's happening, this is the log, fix it". Half the time, it manages to do that.

But even if the fix works, I always retrace it step by step and make sure it's the actual root cause. If the fix doesn't work, however, I almost never iterate with the LLM, other than asking it to find me something in the repo. As you've seen, half of its responses in this case will be very confident hallucinations.

50
1
view more: ‹ prev next ›

AI Coding

55 readers
3 users here now

A community to discuss AI assisted coding.

founded 3 weeks ago
MODERATORS