Nhan Nguyen

On Twitter I like things I don't like.

[Nhan Nguyen @nhan_000]

@Leopeva64 Did you get banned from Reddit?

[Nhan Nguyen @nhan_000]

@GaryMarcus @TaylorLorenz @kashhill I trust @kashhill but I don't think @TaylorLorenz is a trustworthy source. Wasn't she dismissed from @washingtonpost a short while ago?

[Nhan Nguyen @nhan_000]

This really sucks :( I've owned around 7 Pixels in their lineup and am using 3 Pixels simultaneously. I have special love for my Pixel 3 which is running LineageOS and despite the old shell, it feels like a new device.

Really dislike any move that makes the ROM dev's job harder.

[Seang Chau @seangchau]

We're seeing some speculation that AOSP is being discontinued. To be clear, AOSP is NOT going away. AOSP was built on the foundation of being an open platform for device implementations, SoC vendors, and instruction set architectures.

AOSP needs a reference target that is flexible, configurable, and affordable – independent of any particular hardware, including those from Google. For years, developers have been building Cuttlefish (available on GitHub as the reference device for AOSP) and GSI targets from source. We continue to make those available for testing and development purposes.

[Nhan Nguyen @nhan_000]

@seangchau It's been working great the way it is. Cuttlefish can be the reference device, but you don't have to drop support for Pixel devices. Really sour the love I have for my Pixels.

[Nhan Nguyen @nhan_000]

@JohnBar20222395 @GaryMarcus @elonmusk 6-24 or 2 months? https://t.co/TQRcZrPUil

[roon @tszzl]

@DarwinianVyas you guys don’t even know how good you have it. you are within two months of the bleeding edge of model capabilities at all times. there is no government or megacorp on earth with access to better models than you. internet idiots will never grant this to us but openai made ai open

[Nhan Nguyen @nhan_000]

@Charlemagne1 @crest_bell @MJMurthaSr @teka9278 @elonmusk Until you do

[Nhan Nguyen @nhan_000]

@any_other_you However, for 8 GB VRAM I've only heard recommendation up to 8B and not 13B, so I'm quite surprised that the Snapdragon PC can run the 14B model.

[Nhan Nguyen @nhan_000]

@any_other_you I'm quite sure the model size isn't related to the RAM usage, so the 14B model may run, because the Copilot+ Snapdragon PC run these models entirely on its NPU, which apparently is much weaker than a dGPU.

[Nhan Nguyen @nhan_000]

@any_other_you Do you know if it can run DeepSeek R1 distilled 7B and 14B? I've read somewhere that 8 GB VRAM only good for the 1.5B model, but if the Snapdragon PC can run 7 and 14B models I think a dedicated GPU with 8 GB VRAM should be able to run these too.
https://t.co/Y1LEOqMVhq

[Nhan Nguyen @nhan_000]

@any_other_you Will keep Gemma3 4b in mind when I run my first localLLM. Keep putting it off ��. With TTS, I'm running this on Windows. Very good, easy to use, and free too https://t.co/7sEkx83Dgy

[Nhan Nguyen @nhan_000]

@any_other_you Do you think it's the best model one can run locally on 8 GB VRAM (I happen to have the same amount of VRAM as you do 😂). My dream would be to have a decent local LLM + good TTS (I believe Kokoro is new and promising kid on the block) + good STT (Whisper).

[Nhan Nguyen @nhan_000]

@DavidSacks @grok what do you think?

[Nhan Nguyen @nhan_000]

@AndrewFAFO @MPPenaherrera @GuntherEagleman @grok is he retarded?

[Nhan Nguyen @nhan_000]

@Gauldoth_undead @emollick Shhhhhh don't point out the hype!

[Nhan Nguyen @nhan_000]

@ChipOwen6 @LauraLoomer Makes me think a bit

[Nhan Nguyen @nhan_000]

@FFT1776 @grok Is this true?

[Nhan Nguyen @nhan_000]

@elonmusk @grok Is this true?

[Nhan Nguyen @nhan_000]

@GuntherEagleman @grok Is this true?

[Nhan Nguyen @nhan_000]

@9mmsmg @grok Who's this?

[Nhan Nguyen @nhan_000]

@9mmsmg @grok Who's this?