Home Upgrade Search Memberlist Extras Hacker Tools Award Goals Help Wiki Contact

AI Subscriptions (Scam?)

Hey HF,

I'm wondering what is what in this AI landscape. After reading a lot of what php has said about how we can use AI to develop programs, apps and even become this dystopian personal assistant, I tried my best to include it in my day to day workflows.

However, I'm getting very frustrated.. From my own conclusion I thought GPT was no more the standard and this new Claude or Clawd is now the best thing out. I bought their intro price subscription which was around 15 USD (I think) which I then quickly upgraded because within 1 hour of using it to just familiarize my self with their Claude Code app; It kept saying you've reached your limit.

Now this one costs me 25 or something but to my feeling it did not extend my usage at all.. It felt like I was getting rate limited at the same rate. I then bought their Max which is like 100 USD and you guessed it, it rate limited me again. This time not that quick as before, but still way too quick for me to ever get to use it properly.

Besides that, I felt like the whole experience is not in line with all the great things I read and watch about this AI.

[Image: image.png]

I then fired up the AI IDE version of vscode which also has a lot of AI models to pick from to integrate into the IDE but it requires yet another subscription, which i purchased for another 20 USD, I could use it for sometime but within a day I had to upgrade to another tier... upgraded.. then you guessed it, after a day of usage; I had to upgrade another time.

Under the whole influence of it all I spent about +500 bucks on both tools, and the IDE one keeps writing small withdrawals from my bank account.

This is the one that keeps writing from my bank.

When it comes to the AI writing code, it's not that great either; Very weird and unorthodox ways to achieve simple task and also it doesn't obey the prompt.. often goes and does his own implementations resulting in genuinely weird results, often uncalled for code.

I can't help but think this is some form of a scam? I have to wait until XX:XX to be able to continue? Automatic bank withdrawals?

How do you guys go about this?
Yeah bro, AI is powerful, but the limits and subscriptions can be really annoying.
[Image: technology.gif]
TBH I run all via API and maybe spend over 2k a month in credits i guess but thats not all coding stuff media stuff too and more.
But the amount of work I get done with these credits is far more than employing 10 people
[Image: vibesignature.gif]
The Brotherhood Vibe  Legends
(Feb 11, 2026 05:03 AM)7 Wrote:  TBH I run all via API and maybe spend over 2k a month in credits i guess but thats not all coding stuff media stuff too and more.
But the amount of work I get done with these credits is far more than employing 10 people

I was just looking this up, turns out the API service from Claude is different from their subscriptions.. I also didn't realize you can use API keys in the ide.

Does their API do the same kind of rate limiting? I really don't feel like paying to find out. The whole AI IDE thing feels like a expensive gimmick as well, +500 bucks in 2-3 days with practically 0 result.

Don't get me started on it version locking everything to outdated packages either..

I must be doing something wrong?
(Feb 11, 2026 05:48 AM)CL1B Wrote:  I was just looking this up, turns out the API service from Claude is different from their subscriptions.. I also didn't realize you can use API keys in the ide.

Does their API do the same kind of rate limiting? I really don't feel like paying to find out. The whole AI IDE thing feels like a expensive gimmick as well, +500 bucks in 2-3 days with practically 0 result.

Don't get me started on it version locking everything to outdated packages either..

I must be doing something wrong?

tbh I set my bugjets to autotop up so i dont get this issue. Yes I use to have issue of it using outdated packages that why added perplexity as a layer too for realtime updates. Are you paying through the IDE's or each platform and are you talking about claude only?
[Image: vibesignature.gif]
The Brotherhood Vibe  Legends
(Feb 11, 2026 06:05 AM)7 Wrote:  tbh I set my bugjets to autotop up so i dont get this issue. Yes I use to have issue of it using outdated packages that why added perplexity as a layer too for realtime updates. Are you paying through the IDE's or each platform and are you talking about claude only?

Right now I have Chat GPT+, Claude Max subscription (using Claude Code desktop app) and Cursor, the highest one (Using the vscode IDE).

What do you reckon is the best service for human like coding? With human I mean: Write, test, debug and so on.

Edit: And automated
$200 Claude Code = $3000 in inference we're still getting subsidized it's worth it to me

The alternative is using specific models for specific tasks (granular integration) that is optimized and cost controlled.

The final straw is buying a 5090 or hosting on GPU cloud providers on the edge (not sure how well that will work)
(Feb 11, 2026 08:31 AM)php Wrote:  $200 Claude Code = $3000 in inference we're still getting subsidized it's worth it to me

The alternative is using specific models for specific tasks (granular integration) that is optimized and cost controlled.

The final straw is buying a 5090 or hosting on GPU cloud providers on the edge (not sure how well that will work)

This 5090 GPU, is that what it takes to run a custom AI? Besides Claude, Cursor, GPT is there anything else worth looking at? That you know of.
(Feb 11, 2026 08:42 AM)CL1B Wrote:  This 5090 GPU, is that what it takes to run a custom AI? Besides Claude, Cursor, GPT is there anything else worth looking at? That you know of.

Not really, a single 5090 GPU wouldn’t get you the performance you get with claude.

What a bizarre comparison
H4CK3R$ The Academy
(Feb 11, 2026 08:42 AM)CL1B Wrote:  This 5090 GPU, is that what it takes to run a custom AI? Besides Claude, Cursor, GPT is there anything else worth looking at? That you know of.

5090 can run GLM-4.7-flash but nothing compared to SOTA or using a cloud-based B200 or a cluster

Setup serverless and you're cooking
(Feb 11, 2026 10:00 AM)php Wrote:  5090 can run GLM-4.7-flash but nothing compared to SOTA or using a cloud-based B200 or a cluster

Setup serverless and you're cooking

Any references I can read to get started? Don't know all of these models, when it comes to self hosting I only know about Ollama.
Have you looked into Google's Antigravity IDE? All the latest Claude models (including 4.6) are free and available to use inside it.

Antigravity is a fork of VS Code and the models are the "fast" non-thinking versions but its fast and free. Its been fun using that as I learn how to prompt properly and learn to steer it to complete my coding tasks.

I would use Antigravity for most of my tasks coding and if I need the big boy solution I fire up my Claude Code subscription in the terminal to get me un-stuck.
(Feb 11, 2026 11:12 AM)Chinatown Wrote:  Have you looked into Google's Antigravity IDE? All the latest Claude models (including 4.6) are free and available to use inside it.

Antigravity is a fork of VS Code and the models are the "fast" non-thinking versions but its fast and free. Its been fun using that as I learn how to prompt properly and learn to steer it to complete my coding tasks.

I would use Antigravity for most of my tasks coding and if I need the big boy solution I fire up my Claude Code subscription in the terminal to get me un-stuck.

Can you link the Claude Code in the anti gravity terminal? Or what do you mean?
(Feb 11, 2026 11:30 AM)CL1B Wrote:  Can you link the Claude Code in the anti gravity terminal? Or what do you mean?

I haven't tried running Claude Code through the Google IDE terminal. Antigravity boots up with the chat on the right column and I select to use Claude Code Opus 4.6.

Claude then makes changes to my project folder and files to complete my requests I put in the chat.

Its free and simple to use.
(Feb 11, 2026 10:08 AM)CL1B Wrote:  Any references I can read to get started? Don't know all of these models, when it comes to self hosting I only know about Ollama.

Just researching there's no centralized place for this, pick and pull from many sources X / Discord / GitHub / Google / AI deep search