I tend to avoid it like the plague. I don’t find it reliable and it’s half truths and hallucinations end up breaking more than it fixes. Sometimes, I can get a good result, but it’s easier for me to comb forums or carefully google a problem. Transcriptions are awful for us since meetings tend to get really involved so we tell HR not to do AI transcripts.
We have a new AI program that the staff are enjoying and I’m administering creds for it so I may assign myself a user license so I can test it.
That is just the template react vite project. Mine also needed to include a WebGl window so that there was not a canned solution. I find the slightly out of ordinary requests where the AI seems to fail.
I work on gen AI products and am generally tasked with coming up with novel AI tools to try out. I typically use Gemini.
Here are a few things I’ve had AI do recently.
I’m on a large project with many source documents. I add them all to a codeless AI RAG agent and ask it questions about the project primarily for finding commonalities between the docs and retrieving the sources so I can go read them directly.
First round AI code review. It finds a surprising amount of issues. Obviously still need human code review.
Repeated but not exactly formulaic code transformations.
Categorizing sets of data. Define a quick data schema or rating system and have it plough through the data. Always have it explain why.
Opportunistically identify additional unit tests (using a “happy path” unit test as context).
I made an app that helps me with “pebble counts” for stream restoration to determine particle size and distribution so I don’t have to use a pen and paper while standing in a river. Made another for logging my receipts that uses the camera to scan the receipt. Made another that has all the regional curves I use to determine the channel size based on drainage area instead of having to use my excel spreadsheet.
It’s very simple to do. Here’s how:
Tell Chat GPT that you want an ideal first prompt to make a mobile app using Lovable and describe what you want. Chat GPT will give you everything. Review and edit and then copy and paste it into Lovable. Urgent you can easily refine the details with prompts in Lovable. Lovable also easily integrates with Supabase if you need to store data for the app like with a receipt tracker.
I’ve found ChatGPT and Claude to be WAY better than Copilot.
Also try out Perplexity if it’s anything related to finance.
When I am not sure how to accomplish something, I’ll ask ChatGPT “what’s an approach I could take to code XYZ” and use that as a jumping off point. I do the actual programming myself, though. The code it provides is not trustworthy lol. So basically I’m just using it for the kind of things I would have Googled a few years ago. It can save me a few minutes of digging through Stack Overflow or Reddit posts.
Fluval Geomorphologist I guess. Aka Stream and River Restoration. Really I’m just a college drop out who has designed and/or built half a million feet of stream over the last 16 years.
Yes. I’ve done stream work in 12 states from SC to CO. Currently on a project in GA where a bunch of streams were blown up without a permit to install a solar farm and the EPD is making them restore the impacted streams.
I’m the President of the Co. my father is VP and my wife does the books. I used to travel ALOT working on giant projects and had full crew of employees, equipment, etc.. Now with two teens and a one year old, I try to stay closer to home and do smaller jobs within a couple hours of the house.
Awesome, that sounds like so much fun! I’m curious because I have a BS in geology that I never used, except tangentially, as a science teacher. I love hearing about people working in that field, particularly outside of petroleum.
I wrote an iPad application that was for stream restoration for a company called Sitka years ago. It was for use in the field and allowed the user to make notes about stretches of stream with all sorts of statistics and observations. You could take pictures and associate location using GPS.
They demoed it to some people here in Oregon and Washington and everyone loved it. The company eventually decided to make the program work on some stone-age windows device for larger market applicability. I quit because they wanted me to develop using Xamarin which is just horrible. A friend told me later that the users were completely shocked when the application they received was some ascii-based garbage running on windows hand-held devices. Typical government. They do not care what it costs; as long as out looks cheap.
Very cool. I’m a geotechnical engineer in the mining sector, and have worked on a few stream restoration and diversion projects in the past. These days I’m focused pretty heavily on underground mining and tunneling, so the only “streams” I deal with are the ditches in underground drifts that carry inflow to sumps.
I worked in geotech for 6 years as a construction materials tech. Certified to test everything from soil compaction to structural steel bolting. Do not miss those early morning concrete pours and angry superintendents.
It’s extremely rare that I deal with dirt/soil, thankfully. I focus on hard rock pretty much exclusively. I do occasionally need to deal with concrete QA/QC though, mostly the shotcrete used to reinforce underground tunnels.
I have lots of ideas for simple stream restoration apps I’ve always wanted. I just don’t know how to go from having AI code something on Lovable to a product ready for the App Store.
What kind of fee would I be looking at to take an app from online to App Store?
As a cybersecurity engineer I still can’t decide if devs copying from stack over flow or vibe coding is less secure. My gut says both bad but what do I know.
