First enthusiasm A couple of weeks ago I subscribed to Claude Code, and during the first few weeks I had a really nice experience. It was fast, the token allowance was fair, and the quality was good.
I learned they had raised the token allowance for non-rush hours , and since they opposed some governmental rules, it felt good to support the right cause.
(づ ￣ ³￣)づ
However… for about three weeks now my initial enthusiasm has been rapidly waning.
Read in full here:
I honestly look forward to the end of these platforms like Claude. Self-hosting models and using them locally is the way to go, assuming you have the hardware to match it. Plus, there are lighter models like Qwen that can suffice for low-level machines.
In the long run, as expensive computer hardware becomes accessible, Claude Code, Windsurf, Cursor and other companies playing this same token game will end up obsolete.
Beyond that, this obsession with AI for everything just speaks of the luxury of laziness which many developers in lesser parts of the world can’t afford.
Do you have any experience running AI models on your own hardware? If so, can you share your experience, including setup, models, performance, and so on?
How is this man capable of saying what’s in my mind? 
Hi @gfqdjb. I’ve used a low-level Qwen model locally via VSCode and a bunch of extensions like Cline and Continue, but that’s about it.
If you want to go down the same path, I’d recommend this video to help you get started. Beyond that, hardware limitations mean most of the higher models are off limits to me except via the browser or paid API tokens.
I’m on a Mac, an older model, but people with M5 MacBooks still can’t run some of these models.
Let me know if this helps or if you need something else.
Haha thanks, @jaeyson. This AI obsession has me running crazy times! 
Thanks for the info! I’m in a similar position to yours and was trying to figure out how much I’d need to invest to have a productive setup 
I haven’t tried Qwen for coding, but for text analysis, it is good. 
Glad to hear that it helped.
Please save your money. You can be productive with any setup so long as it’s performant.
If AI and LLMs were really that hot to genius level standards, then all these vibe coders would be killing us with super duper amazing technology, but they are NOT.
FYI, there’s a paper that shows a decreasing return in cognitive levels with increased usage of AI. So stay safe and protect your brain 
Oh that’s interesting. What kind of text analysis? Please give some examples. Thanks!
Will it work on macs that’s only 16gb ram?
It should because that’s what I’m using. But pick a light model. Also, be mindful of the storage space. But don’t expect magic; laptops were NOT designed for these models and the limitations are real.
