4 ms·
I think there’s a clear split amongst GenAI developers. One group is consistently trying to play whack-a-mole with different models/tools and prompt engineerin
by ChicagoDave 6mo ago
I think there’s a clear split amongst GenAI developers.
One group is consistently trying to play whack-a-mole with different models/tools and prompt engineering and has shown a sine-wave of success.
The other group, seemingly made up of architects and Domain-Driven Design adherents has had a straight-line of high productivity and generating clean code, regardless of model and tooling.
I have consistently advised all GenAI developers to align with that second group, but it’s clear many developers insist on the whack-a-mole mentality.
I have even wrapped my advice in https://devarch.ai/ https://devarch.ai/ which has codified how I extract a high level of quality code and an ability to manage a complex application.
Anthropic has done some goofy things recently, but they cleaned it up because we all reported issues immediately. I think it’s in their best interests to keep developers happy.
My two cents.
- camel_Snake 6mo agoFYI that prominent link to your sharpee repo on GitHub 404s
- ChicagoDave 6mo agoOMG (fixed) - I updated devarch to check all html links for accuracy. This came from a recent update to the website to simplify it and of course Claude completely hallucinated that URL. You can NEVER stop being vigilant. This is why I still have no faith in things like OpenClaw. Letting an AI just run off unsupervised makes me sweat.
- rglover 6mo agoDead on. Any company not thinking about this like the 2nd group is setting themselves up for a bad time (and sadly, anecdotally, that seems to be an emerging majority).
- ChicagoDave 6mo agoSadly, I agree that the majority are undisciplined.
- joquarky 6mo agoI kind of wonder if people with ADHD tend to fall into the latter group, as we are used to setting guardrails to keep us aligned to a goal.
- ChicagoDave 6mo agoADHD may be a part of it but mentoring and some experience with TDD or strong unit testing is also a part.
- estimator7292 6mo agoIME it seems that output quality is directly proportional to the amount of engineering effort you put in. If a bug happens and you just tell the model to fix it over and over with no critical thinking, you end up with an 800 line shell script meant to change the IP address on an interface (real example). If you stop and engage your brain to reason about bugs and explain the problem, the model can fix it in an acceptable manner. If you want to get good results, you still have to be an engineer about it. The model multiplies the effort you put in. If your effort and input is near zero, you get near zero quality out. If you do the real work and relegate the model to coloring inside the lines, you get excellent results.
- ChicagoDave 6mo agoEven my guardrails can’t replace experience. You have to pay attention. This is exactly how some devs land in whack-a-mole loops.