Remember when AI was $$$ CHEAP?

[joe.pistell]

You’ve used 79% of your Fable 5 limit · Resets Sep 3 at 2:00 AM

This kind of reminds me back in the day when PPC was brand new and we used to pay pennies per click.

Either I'm giving AI much more complex tasks, or they're really throttling back on the free tokens.

 

[Jeff Kershner]

You knew this would eventually happen…

 

[joe.pistell]

Yup, just put another $100 in.

AI will price itself right out. It won't be long before I start to bring this AI in-house.

 

[Alex Snyder]

AI will price itself right out. It won't be long before I start to bring this AI in-house.

You won't get Fable or Opus level capabilities, but Sonnet-level stuff is really excellent for what we are all doing in the car business. We're working with broadly established retail datasets and not trying to fire a rocket around the sun. Totally doable (with a little patience) on your own machine with a downloaded AI.

 

[joe.pistell]

You won't get Fable or Opus level capabilities, but Sonnet-level stuff is really excellent

I hear the Frontier models are ahead of open-source local machines by about six months. And, with the demand for local machines and memory, hardware prices will probably go up. hmmm...

 

[Ryan Everson]

Fable 5.1 just dropped

Much cheaper too - Fable 5.1 High is about the same cost as Fable 5 Low, while performing almost as well as Fable 5 Max $$$

 

[Ryan Everson]

RIP wallet

 

[joe.pistell]

$4100. woa, that is some serious grinding. I do zero dev work, son James is the dev, our 2 man team comes in about $1500-2000 a month.

 

[Burt Lewis]

I think we're still in the very early innings.

I have ChatGPT Plus and, frankly, I'm perfectly fine with it. For what I'm doing, the $20 a month is probably one of the better technology bargains I've ever had.

I'm using AI less like a chatbot and more like another programmer/assistant — building applications, analyzing data, automating repetitive work, creating marketing content, and solving problems that previously would have taken me hours or required another piece of software.

I'm including a screenshot of one page from a dealer dashboard I've been developing. This wasn't some packaged dealer product I bought. It started with conversations with ChatGPT about what I wanted to accomplish. From there, ChatGPT helped me design it, write the code, troubleshoot it, add features, and continually improve it as I come up with new ideas.

That's one of the things I find most interesting about AI-assisted development. The application is never really "finished." I use it, see something I'd like changed or added, discuss it with ChatGPT, and we evolve it another step. Something that started as a fairly simple idea gradually becomes a useful application tailored specifically to how I work and what I want to see.

I've been programming for a long time, and this is probably the biggest productivity change I've seen since the Internet became mainstream. I'm still bringing decades of programming and dealership experience to the table — AI just lets me turn those ideas into working software much faster.

Local AI will definitely have a place, particularly for repetitive dealer-specific tasks where you don't need the latest frontier model. But I don't think it's going to be an either/or decision. We'll probably end up running routine work locally and using the frontier models when we need the horsepower.

The real calculation for me isn't "How much does AI cost?" It's "How much would it have cost me in time, software, or outside help to accomplish the same thing without AI?"

At $20 a month for ChatGPT Plus, that math isn't even close for me.

 

[joe.pistell]

I've been programming for a long time, and this is probably the biggest productivity change I've seen since the Internet became mainstream. I'm still bringing decades of programming and dealership experience to the table — AI just lets me turn those ideas into working software much faster.

@Jeff Kershner and @Alex Snyder
I said this a couple of months ago:

1. AI unlocks the brilliance of men and women inside stores who are deeply immersed in day-to-day life in the dealer world.

2. They will make products and experiences for themselves first, share it with their teams second, and sell it to others third.

3. That being said, based on this simple first-principle observation, DealerRefresh is a beacon for brilliant minds.


DealerRefresh's best days are coming.

p.s. For you young entrepreneurs that are building software systems intended for car dealerships' internal operations, posting open-ended questions like "Hey, do you think this idea will work?", posting here is going to end up as a dead-end question. If you genuinely have the desire to help car dealers build internal systems and help them be better, smarter, faster, then let go of the mouse, get out of the house, and go work for a store or become the general manager's best friend. Build him what he needs.