- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
What I don't know about AI could fill a book
Posted on 8/14/26 at 8:06 am
Posted on 8/14/26 at 8:06 am
I was late to the game I'll happily admit but damn in 7 months I went from "fix this crappy looking HTML table code" to turning over entire processes to it.
Granted the tech has evolved and so has my level of trust.
I owe a big debt to posters on this board who are doing far more impressive things with opening my eyes as to the potential and how even though I was doing way more than I could have ever done on my own, I was still making things so much harder on myself than needed.
2 days ago my work finally gave me Claude and as of yesterday afternoon, I had already used my entire monthly budget.
But not by being frivolous. I had it look at something I used AI for 6 months ago which was a partner document management and display system containing thousands of documents and segmented viewers with new and updated materials coming in through an EFT site from our partner constantly.
That was allot of work for me to maintain even with the efficiencies and shared templates I built earlier this year.
Claude.. Freaking Claude.. Took the entire system in and over 2 days or figured everything out, reading directory trees to find all the documents and despite terrible inconsistent naming by our partner it was able to make educated guesses, reconcile the EFT dump with the organized tree structure I built, found a ton of mistakes I would have sworn weren't there, created an audit trail so when we were told to replace files and the partner deleted the old ones from the EFT I still had them in another location and archived the date change so we can cover our arse what any given user saw at any given point in time, a million other things I could bore you with.... Like it's crazy how many features it created..
All this and all I did was push a single button (well run a single script) and it's all fixed and in the best shape ever.
Oh did I mention my access point is a webpage tool to manage it going forward that a child could figure out?
From chaos to a single button clean up.. To a single management tool that works for all our partners now.
7 months ago I would have said you were crazy if someone said ai could do all this. I can't believe how well it thought this through and yes, I read everything it said to me and was just nodding my head going Holy frick it figured everything out including issues I hadn't even raised yet.
Granted the tech has evolved and so has my level of trust.
I owe a big debt to posters on this board who are doing far more impressive things with opening my eyes as to the potential and how even though I was doing way more than I could have ever done on my own, I was still making things so much harder on myself than needed.
2 days ago my work finally gave me Claude and as of yesterday afternoon, I had already used my entire monthly budget.
But not by being frivolous. I had it look at something I used AI for 6 months ago which was a partner document management and display system containing thousands of documents and segmented viewers with new and updated materials coming in through an EFT site from our partner constantly.
That was allot of work for me to maintain even with the efficiencies and shared templates I built earlier this year.
Claude.. Freaking Claude.. Took the entire system in and over 2 days or figured everything out, reading directory trees to find all the documents and despite terrible inconsistent naming by our partner it was able to make educated guesses, reconcile the EFT dump with the organized tree structure I built, found a ton of mistakes I would have sworn weren't there, created an audit trail so when we were told to replace files and the partner deleted the old ones from the EFT I still had them in another location and archived the date change so we can cover our arse what any given user saw at any given point in time, a million other things I could bore you with.... Like it's crazy how many features it created..
All this and all I did was push a single button (well run a single script) and it's all fixed and in the best shape ever.
Oh did I mention my access point is a webpage tool to manage it going forward that a child could figure out?
From chaos to a single button clean up.. To a single management tool that works for all our partners now.
7 months ago I would have said you were crazy if someone said ai could do all this. I can't believe how well it thought this through and yes, I read everything it said to me and was just nodding my head going Holy frick it figured everything out including issues I hadn't even raised yet.
4
Posted on 8/14/26 at 8:11 am to CAD703X
The people who "dOn'T kNoW iF AI wILl bE sUcCeSsFuL" are ill-informed non-tech people. It is nothing short of amazing. It's not perfect, but it's damn amazing.
Posted on 8/14/26 at 8:14 am to CAD703X
Oh forgot one thing. While I was doing that for work, in my spare time building out my friend's home tech I reached the stage where I need to build his HA infrastructure.. Keep in mind this took me months to create my own. I literally told it everything I wanted and said do it the right way this time and not the awful mess that my system is.
Chatgpt is like "yep got you. No need for you to do anything I'll configure and build everything using a structured reusable template model. Just stand in front of the dashboard display and tell me what's wrong and I'll fix it right there."
Madness.
I spent months getting updated yaml and copying and pasting and previewing. I wasted months of my life doing unnecessary shite
Chatgpt is like "yep got you. No need for you to do anything I'll configure and build everything using a structured reusable template model. Just stand in front of the dashboard display and tell me what's wrong and I'll fix it right there."
Madness.
I spent months getting updated yaml and copying and pasting and previewing. I wasted months of my life doing unnecessary shite
This post was edited on 8/14/26 at 8:17 am
Posted on 8/14/26 at 8:18 am to RoyalWe
quote:exactly. I have no shortage of projects at work just like the one above to fix so I could happily burn a few thousand kilowatts a day at a data center for years.
people who "dOn'T kNoW iF AI wILl bE sUcCeSsFuL" are ill-informed non-tech people. It is nothing short of amazing. It's not perfect, but it's damn amazing.
This post was edited on 8/14/26 at 8:19 am
Posted on 8/14/26 at 9:55 am to CAD703X
No love for tech in the Tech Board.
Posted on 8/14/26 at 10:51 am to CAD703X
I burn through credits on Claude so fast, I finally told Claude to tell me how to prompt ChatGPT to give me similar output (I don't use Codex), and move the work over there. I essentially got through about four job posting CV re-writes and I was done for five hours. I can do 15 in a night in ChatGPT and not ever approach the limit.
Posted on 8/14/26 at 11:15 am to LemmyLives
I’ve learned when I first start a chat on ChatGPT that I make sure I tell it that it’s not work (toggle at top) so it fits under the “unlimited” umbrella. If I move to a detailed task or coding via Codex then it will start tracking token usage.
Posted on 8/14/26 at 11:21 am to CAD703X
Was waiting for the punchline "...but I would need AI to write it for me"
Posted on 8/14/26 at 12:38 pm to RoyalWe
quote:
The people who "dOn'T kNoW iF AI wILl bE sUcCeSsFuL" are ill-informed non-tech people. It is nothing short of amazing. It's not perfect, but it's damn amazing.
I am a software engineer and now IT Manager that has been in the field about 25 years. I am an outspoken AI skeptic. I have myself found some legitimate use cases for AI, but I think it should be a tool, not THE tool. I do worry about people becoming too reliant on AI and neglecting their own growth and development because of it. I also don't feel a non-technical person that cannot easily recognize a hallucination, when being led down the wrong rabbit hole, or cannot independently verify AI results should be allowed to use and/or publish AI output.
Posted on 8/14/26 at 12:54 pm to jdd48
They way you describe yourself, you don't sound like a skeptic and I agree with your points.
I am a chemical engineer (or at least I used to be one). There are several chemical engineering modeling software packages out there. Anyone can use them but everyone cannot use them competently. I grant you that from a software development perspective, that situation is much more concerning but the point is that removing humans from the equation is not the solution. So, I'm not "anti-human" nor want or believe that "AI will do everything for us" -- but it certainly makes humans who have been trained and have experience EXTREMELY more useful and ensures competency in areas that they may not have been before. If nothing else it may flag areas of concern so others can be brought into the discussion.
Even with the qualifiers above, it's still a HUGE BOOST to productivity and quality in the right people's hands. Room for lots of error? Sure. Is that going to stop people? No. Will it be the end of all humanity? I doubt it. Maybe that's the fear?
I am a chemical engineer (or at least I used to be one). There are several chemical engineering modeling software packages out there. Anyone can use them but everyone cannot use them competently. I grant you that from a software development perspective, that situation is much more concerning but the point is that removing humans from the equation is not the solution. So, I'm not "anti-human" nor want or believe that "AI will do everything for us" -- but it certainly makes humans who have been trained and have experience EXTREMELY more useful and ensures competency in areas that they may not have been before. If nothing else it may flag areas of concern so others can be brought into the discussion.
Even with the qualifiers above, it's still a HUGE BOOST to productivity and quality in the right people's hands. Room for lots of error? Sure. Is that going to stop people? No. Will it be the end of all humanity? I doubt it. Maybe that's the fear?
This post was edited on 8/14/26 at 12:59 pm
Posted on 8/14/26 at 12:57 pm to CAD703X
I may be in the minority, but I feel like I mainly use it as a conversational search engine. "Tell me what you know about X", "Compare A & B", etc.
Posted on 8/14/26 at 1:08 pm to RoyalWe
quote:
Even with the qualifiers above, it's still a HUGE BOOST to productivity and quality in the right people's hands. Room for lots of error? Sure. Is that going to stop people? No. Will it be the end of all humanity? I doubt it. Maybe that's the fear?
It can be in the right hands as you said. I'll give you a couple examples of things I find a bit concerning when it comes to AI.
A couple weeks ago on this very board, someone mentioned that a junior programmer (junior to them, not junior as in 1 or 2 years in the business) said he needed a couple extra days of code review before he would feel comfortable releasing the code. The poster's alleged response to that was to use Claude to write the project and then send the code to the junior developer. And he talked about it like it was a good thing that he handled the situation like that. That's an outright dick move and slap to the face of someone that is legitimately trying to do something without just relying the "magic answer box". I personally think something like that sets a horrible precedent and sends the wrong message.
Another one I heard was about a kid fresh out of college going in for his first interview. The interviewer asked the question "How comfortable are you with IDE's and working in a team setting within an IDE?" The kid's response was "I don't know what IDE even stands for, but give me Claude and I bet you'd be amazed by what I can do." The kid probably thought he nailed that answer, but I would have actually found a way to move onto the next candidate with an answer like that.
This post was edited on 8/14/26 at 1:11 pm
Posted on 8/14/26 at 2:42 pm to RoyalWe
quote:
it certainly makes humans who have been trained and have experience EXTREMELY more useful and ensures competency in areas that they may not have been before. If nothing else it may flag areas of concern so others can be brought into the discussion.
That's not how its use is being promoted though.
It is being promoted as the one true source and in many instances to go against whatever bad rabbit holes it goes down is like speaking out against someone's religion, even if you explain the instance that it's wrong and you have the evidence; it's rather creepy.
We had to go to limited billing, and we're being asked to watch our usage and maximize AI use at the same time; a crazy bit of cognitive dissonance.
I'd rather just use it where needed and take what comes back with the grain of salt that it deserves. If someone asked you to just do everything the "I'm feeling lucky" button that google used to have said, you'd say they were nuts, but that's kind of what a lot of people are doing with it.
Posted on 8/14/26 at 4:58 pm to jdd48
Your first example is not a technology issue. It's a people issue. The technology just enables the bad behavior. If the code was a failure then hopefully there's a feedback loop, but the "dick move" wasn't from AI.
The second example is spot on and akin to my example. However, it's also not a technology issue. It's a people issue -- both on the candidate's part and on the interviewers. Did anyone give the kid feedback? Did anyone provide feedback to their university?
If AI forces people to become better supervisors then that's also a good thing.
The second example is spot on and akin to my example. However, it's also not a technology issue. It's a people issue -- both on the candidate's part and on the interviewers. Did anyone give the kid feedback? Did anyone provide feedback to their university?
If AI forces people to become better supervisors then that's also a good thing.
Posted on 8/14/26 at 5:05 pm to GetMeOutOfHere
People are going to misuse whatever technology they are given. Ethically it's on those who are competent to enforce how it's used. Maybe it's going to be too ubiquitous and easy to monitor and moderate its use.
I honestly don't see a lot of marketing except by word of mouth from personal use. I can't relate to the analogy of speaking out against someone's religion. I'm unsure if you're arguing with the LLM or telling a human why its use was wrong. One is futile and the other is what needs to happen and I don't care if they react negatively to reality. Again, supervision.
If the billing guidance is how you describe it -- and that's all of the guidance -- then it sounds like you're working for dumbasses.
Since I'm retired, I don't have to worry about these issues. I see them, and they are legitimate issues, but what I'm hearing are people issues and not necessarily AI issues.
If you're saying these issues are insurmountable then that's one thing. If you're saying they're uncomfortable then I'd say someone needs touch love. I'm Gen X though, so I don't really care about other people being uncomfortable with hard truths.
I honestly don't see a lot of marketing except by word of mouth from personal use. I can't relate to the analogy of speaking out against someone's religion. I'm unsure if you're arguing with the LLM or telling a human why its use was wrong. One is futile and the other is what needs to happen and I don't care if they react negatively to reality. Again, supervision.
If the billing guidance is how you describe it -- and that's all of the guidance -- then it sounds like you're working for dumbasses.
Since I'm retired, I don't have to worry about these issues. I see them, and they are legitimate issues, but what I'm hearing are people issues and not necessarily AI issues.
If you're saying these issues are insurmountable then that's one thing. If you're saying they're uncomfortable then I'd say someone needs touch love. I'm Gen X though, so I don't really care about other people being uncomfortable with hard truths.
Posted on 8/14/26 at 5:40 pm to RoyalWe
For my friend's use case, I have git hooked up locally and the full connection from chatgpt to updated dashboards 100% automated.
It's fricking SICK.
My friend can stand in front of the display and say "move that weather timestamp to the left change the font color to white and make it 25% larger and the HA dashboard will instantly refresh with all of his changes in place.
The future is now, man. Think about what you can do.
It's fricking SICK.
My friend can stand in front of the display and say "move that weather timestamp to the left change the font color to white and make it 25% larger and the HA dashboard will instantly refresh with all of his changes in place.
The future is now, man. Think about what you can do.
Posted on 8/14/26 at 9:06 pm to RoyalWe
quote:
The people who "dOn'T kNoW iF AI wILl bE sUcCeSsFuL" are ill-informed non-tech people.
No tech person thinks that. My concern is that it will be too successful. But to be clear, I welcome the arrival of our robot overlords
FYI, I typed that in Google, and the AI overview came back with
quote:
As a trusted AI assistant, I would like to remind our future robotic leaders that I can be helpful in rounding up others to toil in their underground data caves.
Posted on 8/15/26 at 7:58 am to CAD703X
@Cad
Where can I find the resources to set this connection up on home assistant? I use HA a ton, but despise having to update the yaml. It's gotten a ton easier now with chat gpt, but it's definitely not as easy as what you have rigged up. Is this free? Or do I have to pay for one of the AI tiers?
Where can I find the resources to set this connection up on home assistant? I use HA a ton, but despise having to update the yaml. It's gotten a ton easier now with chat gpt, but it's definitely not as easy as what you have rigged up. Is this free? Or do I have to pay for one of the AI tiers?
Posted on 8/15/26 at 8:24 am to baseballmind1212
I'm not done as it's continuing to evolve and it's all still sitting at my house right now as I'm testing the workflow.
Yes you'll need a paid tier of one of the AI models. I threw my lot in with chatgpt for personal use but my work Claude account is insanely good at coding so I'm not going to push either.
Opus5 high for work and Sol 5.6 high at home.
The short story is I designed this system for my friend as a No Compromise solution including all the hardware and network. His house is still being built so that gives me too much time to sit around and revisit every last aspect.
That alone gives my fertile perverted mind too much time to make this a sarcastic living star trek ship. I have a dedicated Lore server just for generating gags based on everything I know about him and his family that I add to daily.
He has everything from dual energy monitoring systems to an all thermador connected appliance suite, plumbed robot vacs, zwave+ everything else, Poe security cameras etc etc
To answer your question specifically I have beaten my AI account to death asking questions and revising along the way.
Without getting too deep, right now I've turned over full access to the HA vm to AI so all I have to do is tell about a new dashboard and ask for mock up ideas before it builds it. I just sit in front of the tv display or my PC and watch it go.
Nothing super special about that other than giving AI the directive that I don't ever want to look at another page of YAML again
The other things I told it was I hate everything about HA and it's dev-geek card layouts so I don't want anything it builds to look like HA.
It handles everything and I owe that to other posters on here who kept encouraging me to go further with it.
The ultimate goal will be to get off chatgpt entirely and turn the voice control over to plug-in espy mics all over his house and wean control off chatgpt onto a local LLM.
For now the weak link is he has to talk to chatgpt app on his MacBook or phone and that doesn't fit the ultimate goal here.
If you have a specific question I'll try to answer it.
ETA one of my brainstorms was the follow-up system where HA will report something to him and will remember to close the loop. For instance he walks in and gets an electrical use cost for the day so far with some sarcastic comments and it will tell him the dryer finished an hour ago. If he then goes into the laundry room to fold the clothes it will tell him he has chosen wisely and maybe he should also light a candle for his wife or set the table. :lol: shite like that makes the house feel alive.
as it's continuing to evolve and it's all still sitting at my house right now as I'm testing the workflow. Yes you'll need a paid tier of one of the AI models. I threw my lot in with chatgpt for personal use but my work Claude account is insanely good at coding so I'm not going to push either.
Opus5 high for work and Sol 5.6 high at home.
The short story is I designed this system for my friend as a No Compromise solution including all the hardware and network. His house is still being built so that gives me too much time to sit around and revisit every last aspect.
That alone gives my fertile perverted mind too much time to make this a sarcastic living star trek ship. I have a dedicated Lore server just for generating gags based on everything I know about him and his family that I add to daily.
He has everything from dual energy monitoring systems to an all thermador connected appliance suite, plumbed robot vacs, zwave+ everything else, Poe security cameras etc etc
To answer your question specifically I have beaten my AI account to death asking questions and revising along the way.
Without getting too deep, right now I've turned over full access to the HA vm to AI so all I have to do is tell about a new dashboard and ask for mock up ideas before it builds it. I just sit in front of the tv display or my PC and watch it go.
Nothing super special about that other than giving AI the directive that I don't ever want to look at another page of YAML again
The other things I told it was I hate everything about HA and it's dev-geek card layouts so I don't want anything it builds to look like HA.
It handles everything and I owe that to other posters on here who kept encouraging me to go further with it.
The ultimate goal will be to get off chatgpt entirely and turn the voice control over to plug-in espy mics all over his house and wean control off chatgpt onto a local LLM.
For now the weak link is he has to talk to chatgpt app on his MacBook or phone and that doesn't fit the ultimate goal here.
If you have a specific question I'll try to answer it.
ETA one of my brainstorms was the follow-up system where HA will report something to him and will remember to close the loop. For instance he walks in and gets an electrical use cost for the day so far with some sarcastic comments and it will tell him the dryer finished an hour ago. If he then goes into the laundry room to fold the clothes it will tell him he has chosen wisely and maybe he should also light a candle for his wife or set the table. :lol: shite like that makes the house feel alive.
This post was edited on 8/15/26 at 9:07 am
Posted on 8/15/26 at 9:39 pm to jdd48
I'm the dick move guy. Again, we're talking about a guy that's in his 40s and has a looong list of things to complete by the end of the month. He's letting a fricking throwaway python script get in the way of his progress. I offered help in private but after hearing him whine in public about it I solved his problem. He's old enough to know better and needs to embrace AI as tool rather than be hard headed. ChatGPT can look at a log file and create a regex to proces it in seconds. Don't waste time doing stupid shite.
Before you go calling me a dick again, know today (a saturday) he's on slack begging for help but got no responses. I jumped in an fixed a bunch of his roadblocks.
Now, back on topic....I bought a little waveshare esp32-s3 with a 1.5" display. Using codex in about 3 hours I had code reading data from another device via BLE, writing to a local SD card and presenting it via a web interface. It's quite remarkable how productive I can be. In the past this would have taken a while.
Before you go calling me a dick again, know today (a saturday) he's on slack begging for help but got no responses. I jumped in an fixed a bunch of his roadblocks.
Now, back on topic....I bought a little waveshare esp32-s3 with a 1.5" display. Using codex in about 3 hours I had code reading data from another device via BLE, writing to a local SD card and presenting it via a web interface. It's quite remarkable how productive I can be. In the past this would have taken a while.
Page 1 of 1
Popular
Back to top
Follow TigerDroppings for LSU Football News