Mir Mujtaba

Helping Dentists get Patients through Leveraged Marketing. 7+ Clients served. YouTube : ads, Content + MORE
👇

[Mir Mujtaba @notmirmujtaba]

Want to never worry about getting patients?

Get my free marketing audit + strategy + free call(deep growth plan)

And get more patient flow in your practice.

Dm me “free” to get it

[Mir Mujtaba @notmirmujtaba]

@0xSero We are almost the same height btw, I’m 5,5.5

[Mir Mujtaba @notmirmujtaba]

@AlexFinn Now tides have changed, Gemma 4-12b can do most of the agentic work and with Hermes can figure out how to do it if it doesn’t know how. Long term my bets are with Google Open source though, QWEN is becoming a closed source and that worries me a lot

[Mir Mujtaba @notmirmujtaba]

@sharbel Mac mini can run Gemma 12B QAT with ease and good context, meaning agentic tasks, normal workflows can be done locally. For high level tasks we should be making a dumb-proof workflow guide/skill from the frontier model and feed it to Gemma/Qwen/LFM ect (for 16GB ram users)

[Mir Mujtaba @notmirmujtaba]

@0xSero Google is doing more for open source than anybody thought they will do, fr I’m surprised…

[Mir Mujtaba @notmirmujtaba]

@0xSero https://t.co/tZrNzr5cgZ purchase it, make a small group of people with the best knowledge and hardware, The rest can give their computing power (like servers). Then you already have a small team and people lending their compute to compress, create, gather more models/data.

[Mir Mujtaba @notmirmujtaba]

@JackdeS11 Context?

[Mir Mujtaba @notmirmujtaba]

@PirouzOfPersia @RyanRozbiani They are zios with a different name in my opinion.

[Seyed Abbas Araghchi @araghchi]

Foreign forces in proximity to our territory are at constant risk on account of their own human errors, plain accidents, or potentially being caught in crossfire.

To reduce risk, best solution is for them to leave.

We prefer language of diplomacy but speak other languages too.

[Mir Mujtaba @notmirmujtaba]

@cjzafir No open source?

[Mir Mujtaba @notmirmujtaba]

@theo What happened to TPU’s?

[Mir Mujtaba @notmirmujtaba]

@cjzafir Is this real?

[Mir Mujtaba @notmirmujtaba]

@0xSero bro skipped 16-32-64 range like its nothing....

[Mir Mujtaba @notmirmujtaba]

@0xSero Not trying because I know my luck is shit even typing this

[Mir Mujtaba @notmirmujtaba]

@NVIDIAAI Please let him know about me 😂

[Mir Mujtaba @notmirmujtaba]

@luigifcruz @NVIDIAAI @BerkeleySETI Please know me man

[Mir Mujtaba @notmirmujtaba]

@jun_song can 1 bit models on cpu then gpu solve this? prism llm proves it is possible, just need google to make gemma 5 with a 1-bit version

[Mir Mujtaba @notmirmujtaba]

I RAN A QWEN 3.6 35B-A3B (quant version) on a RTX 3050 with FREAKING 4gb VRAM! And ONLY 16GB CPU RAM.

At 64k Context, about 5-6TPS….

[Mir Mujtaba @notmirmujtaba]

@ai_hakase_ Can it run on RTX 3050 (4gb vram) and 16GB ram?

[Mir Mujtaba @notmirmujtaba]

@above_spec can this run on RTX 3050 (4GB VRAM) and RYZEN 7 (16GB RAM)??
please tell me!

[Mir Mujtaba @notmirmujtaba]

@KyleHessling1 Thanks man, will try this today!