Used Chat GPT today to help me figure out how to import .tiff and .shp files into civil 3D then create a shaded banding topo map for an Army Corps of Engineers permitting document. Would have probably never figured this out on my own. I don’t see AI as thinking for me, instead I’m learning new things I otherwise wouldn’t.
2 Likes
I refactored a library to build multiple versions of our component library using AI. What should have taken all afternoon was done in an hour. Simple, but time-consuming. That is AI aiding my productivity for sure. Now, if I could only get Storybook to stop releasing all these breaking changes; I would be most happy.
I am not as excited about it writing actual code, but using it for simple things is almost as useful.
1 Like
We have a new AI based IDE that is mandatory but not mandatory. It has an AI buddy in it that has the personality of Clippy crossed with a golden retriever on 8 cups of coffee.
I asked it for help upgrading some dependencies to the latest versions and resolving breaking changes. It’s connected to our repo and supposed to be able to pull info on packages but instead of looking up the real latest versions of packages it iterated every single dependency to the nearest whole number. IE Spring Frame 7.0, Python 4.0, JDK 26. No matter how imaginary the release was. It cheerfully announced that it was finished and that the new branch will absolutely build and run perfectly. I pointed out how absolutely wrong it was and it complimented me on my astute observation and then created and ran a series of 8 bash scripts to try to undue its work instead of simply stashing the local changes.
It wasn’t helpful but it sure was entertaining.
5 Likes
I have been using Claude all this week to build a parts list generator from inspection reports for my company. It’s absolutely insane what I was able to accomplish, I’m blown away by what it can build with enough guidance.
We also have was that we have a diag agent trained in all our manuals. It’s pretty accurate, but it makes mistakes here and there and there’s no replacement for field experience for sure.
2 Likes
This is what I think is the value of AI right now. The ROI on the application you describe would not be justifiable to hire a human/professional to write. But the knowledge of what you want and need allows you to create this function easily and cheaply using AI and a bit of your own brain power working with the AI to get things right.
Bravo.
I created a new set of actions for a page-view to store the URLSearchParameters for retrieval on navigation for certain application pathways. Took a few minutes to write the description and outline the paths to the relevant files.
The tool is definitely evolving. The AI is getting better too 
I deliberately avoid Ai at work (as I also avoid in general everyday life). I’m a graphic artist, I will create my own stuff.
7 Likes
AI art is a blight on humanity
8 Likes
I am pretty sure this is how all new art movements are described at the time of inception.
1 Like
Ya got me there, old man!
1 Like
If anyone can create AI art it will always be inherently worthless.
3 Likes
Well currently unemployed and definitely using ai for job search and such
5 Likes
Oh no, sorry to hear man, wish you the best in your search
3 Likes
100% agree on the hallucination problem, especially with interns. They trust the autocomplete blindly and end up spending hours debugging functions that don’t even exist in the docs.
We had the same issue on our team last year. Copilot was okay for writing quick unit tests, but terrible at understanding our actual codebase context.
We ended up adding Sugarbug into our GitHub workflow just to catch those hallucinated imports and broken syntax before PRs even reach review. Saves a ton of headache.
It won’t replace devs anytime soon, but it works alright if you treat it as a glorified linter rather than an engineer.
What models have you tried?
The biggest problem I have been running into is the inability of AI to understand business rules and intent. To get meaningful results, I need to break the problem down with prompts and context-hints to a large degree with trial and error. By the time I get something I can use; I could have just done it myself; correctly, the first time.
Then there is the trust issue. Don I trust AI to understand all of the ramifications of the changes it is attempting to make to complex code? The answer is NO!
With some preparation you can build a framework that will guide the AI on business rules and such. It will take some work but it’s basically a way to prep automation. Do it once then use it repeatedly. Depending on what you want it to follow and such you would have to do some research and test different things out until you found what worked but you can do it. Trust is another thing, you definitely have to read every line of code and chew the agent out sometimes. I’ve had folks who don’t know how to code make stuff with AI 
fixing their stuff using AI and proper frameworks is a nightmare and often it’s better to throw their work out.
This is coming from someone who has a huge list of rules and styles and stuff to follow and hasn’t written a single line of code in months as a developer. I’ve had to take the reins and debug sometimes, but even then I don’t write the fixes. It’s corporate mandated so I had no choice but to figure it out if I wanted to not get PIP’ed.
Model. Claude 4.6 Opus is my preference. They just took that away, I’m really sad about it. Fable 5 and stuff is a downgrade imo. It argues with me and is less likely to follow those structures and rules. Oh I read your instructions but I knew better so I did this thing that is technically more efficient but goes against company policies.
Since my company does not have any mandates, I wonder if using AI is:
- More efficient
- Faster
- Easier
- or, just different.
Depends on the situation and what you are doing. In my experience it is faster, more efficient, and is a significantly worse drain on your mental health compared to traditional coding. 
I get so much more done but have to argue with an imaginary toddler that is really good at coding but an idiot.
2 Likes
Yes, using it. Love/hate relationship. Can’t beat them, join them.. efficiency gains cannot be ignored.. etc.
I find it works really good if you adhere to KISS principles and don’t ask it to engineer something complex in one go. I break projects down into small tasks and then leverage AI to build those tasks. Things usually come together pretty well.
I regret investing so much of my time/education in computer science. I know my working days are numbered. I wish we were not careening toward skynet.
I’d rather be a cabinet maker or something.
3 Likes
Yes that is my experience also. It is just that some tasks are so complex; that breaking it down so AI can understand them can make more work than just doing the task myself. Right now I am implementing a dynamic UI. Trying to describe what I want and what I want it to do would be more difficult than just doing it. Without amazon mandating I use AI; it is always a bit of a call as to what course to take.
I imagine that my hybrid approach is likely less efficient than just going all-in on one or the other.
1 Like