Tech Thread
- Thread starter MindzEye
- Start date
MindzEye
Wayward Ditch Pig
I'm not remotely surprised. I see the current value in a Claude Pro subscription for a small business person, at the current rates (which are heavily subsidized by investor money), but paying actual token rates for a large business is motherfucking crazy. When they switched to token based billing a while back big companies like Uber were ripping through their annual budgeted spending threshold in 4-6 weeks. This is another reason that the industry is a dead man walking. How do you sell a product that is 40-50x more expensive than competing open source models to run, that is only marginally better at the things that LLM's are actually good at. Throwing a shit ton of tokens for improved "reasoning" makes the model easier to work with as it understands less precise prompting better, is better at looping back on itself to troubleshoot stuff it's working on before providing the output, etc. But that's a quality of life improvement for the end user, and doesn't noticeably improve the end product for most things.
CH1
The Artist Formerly Known as chiggins.
I give Claude and Chat a combined $40 a month of my money every month.
Don’t really have a A to Z use case right now but honestly in my field if I don’t know how to use AI, I’ll be fucked.
I find it extremely valuable for quick prototyping. I can mock up something fast and show stakeholders what something will look, feel, and sound like before they commit more money for me to deliver a great final product.
Don’t really have a A to Z use case right now but honestly in my field if I don’t know how to use AI, I’ll be fucked.
I find it extremely valuable for quick prototyping. I can mock up something fast and show stakeholders what something will look, feel, and sound like before they commit more money for me to deliver a great final product.
MindzEye
Wayward Ditch Pig
This is where it's smart to separate the technology from the industry though. If the industry collapses, building an instance of Deepseek to run on a dedicated machine isn't outside of my personal technical skills and it would have most of the capabilities of the current Claude. Probably just as good at writing code, probably less good at "understanding" my prompts when I'm explaining what I need it to do. Shit, there are communities online that are building custom harnesses that access the specific LLM only for what they're good at, and run the grunt work through Deepseek at small fraction of the cost (looks like current token rates are 0.14 for 1 million tokens on Deepseek, $10 per million tokens on Claude.
As far as pure affordability goes, the capability of deepseek is where the industry really is. It's not going anywhere, it's a sustainable product.
LeafOfFaith
Well-known member
I spend $20 a month on Chat and use free Gemini here and there. I find it an incredibly helpful tool in my work. Something to bounce ideas off of, or check things against, or to draft basic provisions here and there so that I don't have to search through prior deals to find the right starting language. I legit can't believe what it's capable of. I still prefer my own voice and structure in letter writing, unless it's a bullshit letter for something that I don't care about. For example, client's wife had someone hack into her Apply pay and spend a few grand and couldn't get them to reverse it, so they asked me to help as an unpaid favor. One prompt and a letter generated in two seconds that I sent off and they reversed it the next day - normally I'd have to spend a couple of hours with research and writing, now two seconds instead of two unpaid hours and I'm a hero to the client).
Some people say you HAVE to get the premium Chat version, and I'm like, what more could it possibly do? I almost don't want to know.
