Alex Ziskind
13.4K posts
Alex Ziskind
@digitalix
Developer and content creator, founder of @NativeScripting, ”the voice of NativeScript”, trainer, author, speaker, synth tweaker. https://t.co/ZjhwZDX24l
Washington DC, USA Katılım Mart 2009
1.5K Takip Edilen21.7K Takipçiler
@digitalix Did my part from 2nd account also, u already reached 500k bro, congratulations !! 🥳
Video that hooked me to your channel
English
@digitalix I hear that 1M is a number after which the mouth starts slowly closing back in on the thumbnails, I'm sure you'll be there in no time :)
Jokes aside, congrats on the milestone, making local inference accessible is probably one of the most important things these days!
English
Alex Ziskind retweetledi
Early look at Etudera, my interactive music practice app. Built with GPT 5.5 in Codex.
Synced notation + guitar tabs, playback, tempo control, looping, metronome, MusicXML import, and dark mode.
Still early, but it’s starting to feel real. 🎸
English
@digitalix I am spearheading the development of a transformative containerized AI orchestration platform that empowers 10,000+ developers worldwide to deploy enterprise-grade local LLM infrastructure.
yep
English
@ivanfioravanti picking one or two is a hard problem. you’ll get 20 people with 20 different answers here
English
In a Local AI benchmark what should be the context length used as main reference for speed?
We are all using very small context to showcase model speed in our hardware, but that's unrealistic.
English
Looks like you cruising the aisles at MC..."I'll take two of those...five of those...". Writing of, just got back from MC Charlotte with an Open-Box 16" MB Pro M5 Max w/64GB, max cores, 2GB SSD - $4.8K OTD...but what delivers tomorrow from B&H Photo is the big news = 2 x M3 Ultra "Big Boys!". HUZZA!!!
English
Gemma is so much faster! wait, are the ghosts walking through walls? did they idspispopd?
tawer@tawer1O
Comparison of Gemma 4 and Qwen 3.6 in the same coding task Same hardware, same prompt, comparable model size > gemma 4 31b: 27 tok/s, 3m 51s, 6,209 tokens, stronger game logic > qwen 3.6 27b: 32 tok/s, 18m 04s, 33,946 tokens, better visuals Gemma 4 with lower tok/s finished 14 minutes faster because it used 5.5x fewer tokens to reach a complete answer
English
Who said GPT 5.5 can't do front-end? It's coming along great, I think.
English
Interview I did for my GPT-3 social robot 3 years ago, I have never shared this on here
Matt Shearer@MattWBZ
Faces of the future. @Ominousind @wbznewsradio
English
@digitalix I used to really enjoy your videos... I think you have fallen to the YT fever, however.
English
I split one LLM across a DGX Spark and a M3 Ultra Mac Studio to see if they would be faster together (new video)
English
@Authentic1ty @erictimmer asus has extra heat pipes if i remember correctly
English
@digitalix @erictimmer ASUS also cools the best. The physical unit is larger and has better ventilation.
English
@digitalix Worth reaching out to them? Idk of another 16x spark cluster but would be happy to contribute to Exo and run tests on this
English
@Kurcide pretty slick cluster you have there. i recommend waiting until the official support is out from exo, which should be pretty soon
English
@digitalix Just built a 16x DGX Spark cluster and want to test pairing with Mac studios for decode.
Can we connect?
English
@digitalix Yep! Keep up the great content, watched a lot of your videos and ended up buying the machines so I could do local. Glad I could contribute back.
English
Watch @digitalix newest video to see how you can set this up too.
Thanh Pham@runsonai
4 hours of grinding. Running on untested branches to make nvidiva gb10 chip work with mac studio on the same cluster. But I got it working now @exolabs! Running Qwen 3.5 27b, 2100+ t/s prefill speed on my gx10, 19 t/s on studio, 4200ms ttft. Not bad before optimization.
English