Anyone here played with AI agents and can recommend one ?

[Deschodt]

I want to setup a simple AI agent that, based on knowing when my kids school schedule (different every day) , will look up the fastest route from my house (traffic) and email me a recommendation 5 min before I go. Pretty simple stuff but useful..

Many tools out there.. I suppose I could ask AI itself but it could be biased ;-) Anyone recommend a toolset, preferably free ?

[Paul T]

I have VERY little experience in this area but am trying to learn a little. Check out Claude by Anthropic - I've messed around with it a little and have been impressed as someone who doesn't know what they are doing.

[masraum]

I'm not familiar with the agents. I heard about them for the first time today, and will be learning more about them next week. I'm going to try to embrace this whole AI thing since it's the future and if I don't get a handle on it, that'll be limiting in my job vs enhancing.

From what I understand, they can be extraordinarily useful, so I'm curious to see where this thread goes.

I know a guy that said that he wanted to take a bunch of old DV tapes and digitize them for use with computers. He said that there was stuff to do it, but that it was kind of expensive. He used an AI agent that created an app to extract them to a computer.

[Deschodt]

I spent a few hours looking and so far, unless it's something you really wanna get into for work, it's the wild west.. I narrowed it down to some apps (claude code, cursor AI, make.com) but ultimately, while I do have the concept on paper and could code it +/- (query maps at certain times, give me the fastest route, send it to me), the most basic functions require that you integrate a bunch of platforms and accounts together (the agent, the target, the AI itself, then the output) - it's a gigantic pain in the arse... Google maps API alone require a google cloud account and "billing info" to get an API key that you can then feed into Make.com and then same withClaude or Chat GPT, and then back to Gmail.. Uf, I thought this was more user friendly by now, I will wait till someone makes a well integrated platform... Or maybe that exists but it's behind a paywall... I won't be ahead of that curve..

[Paul T]

I’ve messed around with Replit a little also, might be worth a look…

[rcooled]

Originally Posted by Deschodt

...will look up the fastest route from my house (traffic) and email me a recommendation 5 min before I go.

Saving the school's location in your phone's maps and checking for the fastest route before you go won't work?

[RANDY P]

Waze?

[Alan A]

this isn't an easy problem - without going into the weeds and actually coding something.
there is an mcp server that links waze and claude.
looking into that to start is probably going to be your best bet.

[stealthn]

AI is garbage and dangerous, you are the content…

[Paul T]

Originally Posted by stealthn

AI is garbage and dangerous, you are the content…

I disagree with that as a blanket statement….there is a LOT more to AI than just the generic LLM’s the most people are familiar with (Chat GBT, etc). I agree that those are largely just pattern recognition engines that frequently get things wrong (and confidently so!), but there are all sorts of emerging technologies that show huge promise in any number of fields. I think the next 5 years are going to be amazing. Ignore it at your peril, especially if you are an investor or just like staying on top of current tech. It would be akin to claiming the internet was useless in 1985.

[stealthn]

I get it, each to their own. Trust me I am not ignoring it, I am trying to help companies protect themselves from the negatives.

The one big thing that is needed for it to succeed is universal guard rails, sadly due to $$ and politics I don't see this happening.

[Shaun @ Tru6]

I wanted to use an AI agent to help me finish my screenplay but immediately realized my content will be shared with the entire world in one keystroke.

AI will ultimately be extremely detrimental to 60 to 80% of the population and only help the wealthy become wealthier while everyone pays higher electric rates to make them richer while losing their jobs. Win win.

[Cajundaddy]

Call me foolish but I don't want to train AI on the location of my kids every day. Everything we enter trains them to learn from us and some things are better left unlearned IMHO.

I set alerts in my phone and use Waze for the most efficient routing in traffic. A few keystrokes seems like a small price for kid location privacy.

[3rd_gear_Ted]

Life 360 is what my family uses to track each others whereabouts.

[Paul T]

Originally Posted by Shaun @ Tru6

I wanted to use an AI agent to help me finish my screenplay but immediately realized my content will be shared with the entire world in one keystroke.

AI will ultimately be extremely detrimental to 60 to 80% of the population and only help the wealthy become wealthier while everyone pays higher electric rates to make them richer while losing their jobs. Win win.

I’m not quite that bearish on it, but it will most definitely have a net detrimental impact on jobs, I would imagine. I think with most professions though, it will come to be used as just another tool, rather than an outright job replacement.

Now, if you want to get real conspiratorial, there are those that think AI was created simply to preserve human intelligence to be used in the aftermath of the next geophysical mass casualty event, which the “elites” know is coming soon, which will destroy ~95% of life on earth. Just to brighten your day, lol….

[masraum]

Originally Posted by Paul T

I’m not quite that bearish on it, but it will most definitely have a net detrimental impact on jobs, I would imagine. I think with most professions though, it will come to be used as just another tool, rather than an outright job replacement.

Now, if you want to get real conspiratorial, there are those that think AI was created simply to preserve human intelligence to be used in the aftermath of the next geophysical mass casualty event, which the “elites” know is coming soon, which will destroy ~95% of life on earth. Just to brighten your day, lol….

My guess is that certain sectors will see some reductions. If there's a team of 15 people doing a certain job at a certain company today, then tomorrow or 3 or 6 or 10 years from now, there will be 5 or maybe 2 or 3 people doing the same job more efficiently than the team of 15 people doing it today.

And the folks that are great at a particular job today, may not be the best at it when it's heavily using AI in 5-10 years.

Many years ago, I managed a FLAPs (for a year). We had a cashier that I realize now was probably on the spectrum, but functional. When I started, our cash register was an old style cash register. She was constantly making mistakes, often the same mistakes over and over again never learning even when I explained what she was doing wrong. The company changed to a computerized point of sale system. I dreaded this girl using that system. She took to it like she'd written the program herself. She never made a mistake, and usually fixed other peoples' issues.

I work with 2 guys now. Both are network engineers. One is strongly resisting learning anything to do with automation. The other has embraced it and is using agents both at work and at home to create things, and he's very successful at it. He has said "I'm a network guy not a coder, but if you came to my desk, I'm embarrassed to say that I'd probably look like a coder." (I'm not sure why that would be embarrassing). I'm hoping to learn from him.

I've tried coding off and on over the years. I remember Basic when I was a kid and we had a C64. Coding for days out of a magazine or book for nothing useful to happen or for a typo on some line out of 100,000 lines to be causing it to not work. Ugh. And then Fortran 77 in college, double ugh!

I've now done a little bash scripting and sql and a little bit of other stuff. But some of this stuff (when it makes sense and has a practical application that seems useful) is great. I'd love to learn to do this stuff, and I'm going to give it a go. Of course for me, it seems like a smart career move if I want to be able to work until I'm ready to retire.

But yes, there are lots of holes, HUGE HOLES today that are a bit scary that most folks don't think about, aren't aware of, and aren't scared by. I don't intend to share everything with the world and make myself a commodity (more than we already are).

[Alan A]

You all need to look at the T&Cs of commercial (paid for) AIs vs the free ones.

You think a fortune 500 wants (allows) their data to be used to train an LLM and it be in the wild?

[flipper35]

Originally Posted by Alan A

You all need to look at the T&Cs of commercial (paid for) AIs vs the free ones.

You think a fortune 500 wants (allows) their data to be used to train an LLM and it be in the wild?

The AI companies don't care about the T&C where they get their data, routinely scraping without permission or torrenting books and news articles. I don't know any that are on the up and up when it comes to that. Some are just less shady than others.

For the OP:

Waze will do it based on time of day and you can set departure times so you have real time traffic updates.

Not that I really like Waze, but it does that well enough.

[Alan A]

I’m not talking about scraped - public - data. I’m talking about IP.

All of the AI companies with commercial offerings explicitly exclude - or allow the user to exclude - their data from retention and use in training. They wouldn’t have a commercial offering without same.

[unclebilly]

If you have the ChatGPT app on your phone, it remembers your chats. I suspect you could call up this same chat over and over and ask it to recalculate every time. Not perfect but it would work… then again, Apple Maps and Google maps can do this too.