Alex Volkov @ AI Engineer (@altryne) / X

Alex Volkov @ AI Engineer

54.3K posts

Alex Volkov @ AI Engineer

@altryne

🎙️ Host of @thursdai_pod ✨ AI Evangelist with @wandb 🪄🐝 working on @weave_wb (opinions my own)

Pod Newsletter →

Joined May 2007

Pinned
Alex Volkov @ AI Engineer
@altryne
Jul 11
We moved from 3 frontier AI labs to 5 in a single week! This week was the craziest week in AI news, almost since we've done the @thursdai_pod - new episode and a 5500 word newsletter is here! thursdai.news/jul-9 We covered about 9 releases, tested 5 of them on air, had
📅 ThursdAI - live 8:30am every Thu
@thursdai_pod
Jul 10
Replying to @thursdai_pod
The full write-up, every link and benchmark: thursdai.news/jul-9 Interactive episode page with the four-model Infographic Arena: thursdai.news/ep/jul-09-2026
Alex Volkov @ AI Engineer
@altryne
Aug 5, 2024
I'm struggling with this, so let's talk about it. Google drove vehicles across all of US (and the world) and took pictures of everyone and everything, yeah they blurred faces, but nobody gave them permission, and now we have Google Street View. How isn't this the same?
Marques Brownlee
@MKBHD
Aug 5, 2024
cool cool cool cool cool cool now leaked NVIDIA slack messages discussing which YouTube channels to scrape videos from. MKBHD videos? Yeah grab those too.
Alex Volkov @ AI Engineer
@altryne
Nov 6, 2024
Replying to @sama
Alex Volkov @ AI Engineer
@altryne
Apr 21, 2025
HOLY CRAP, a new super tiny 1.6B param voice model just dropped that seems to.. outperform 11labs!? 😵‍💫 From Nari-labs, Dia is an Apache 2.0 voice model, that can generate laughs, sniffs and emotions, copy an existing voice and is effectively real time on larger GPUs:
00:00
Alex Volkov @ AI Engineer
@altryne
Jun 9, 2025
Frontier Labs: Our AIs are near PHD level and will do most of your work by end of year Apple:
00:00
Alex Volkov @ AI Engineer
@altryne
Nov 20, 2024
Chat, is this better than having your camera turned off? (for $60/mo) x.com/getpickleai/st…
Alex Volkov @ AI Engineer
@altryne
Jan 20, 2025
POV: A chinese company drops a powerful open source reasoning AI model competitive with o1 that neither of your companies have released yet as products while you're attending an inauguration
Alex Volkov @ AI Engineer
@altryne
May 24, 2023
Watching @karpathy presentation from today and taking twitter notes, come along for the ride: If you're like only the practical tips, skip to #32 @karpathy starts with stages: 1 - Pre-training - months x thousands of GPUs 2, 3, 4 - Finetuning stages that take hours or days 1/
Alex Volkov @ AI Engineer
@altryne
Feb 3, 2025
Consultancy firms are so fucked DeepResearch FTW
00:00
Alex Volkov @ AI Engineer
@altryne
Dec 4, 2024
Google just... broke SOTA weather prediction with a... diffusion model 😮 or rather, an ensemble of 50 or more predictions and open sourced it! Beating the current SOTA called ENS at 97% of 1320 combinations of predictions and 99% of lead times of greater than 36 hours.
Google DeepMind
@GoogleDeepMind
Dec 4, 2024
Today in @Nature, we’re presenting GenCast: our new AI weather model which gives us the probabilities of different weather conditions up to 15 days ahead with state-of-the-art accuracy. ☁️⚡ Here’s how the technology works. 🧵goo.gle/49trAOv
Alex Volkov @ AI Engineer
@altryne
Dec 2, 2022
I turned #chatGPT into a macOS status bar app, with a shortcut (⌘+ctrl+G) and the chat bar even autofocuses. So now all this power, is always at my fingertips
Alex Volkov @ AI Engineer
@altryne
Nov 19, 2023
Incredible to see, I have a list of OpenAI folks on X, I've never seen anything like this ->
00:00
This post is unavailable.
Alex Volkov @ AI Engineer
@altryne
Dec 3, 2022
My @OpenAI #chatGPT is not like your chatGPT. Mine can... draw 🤪 (using @StabilityAI API or Dalle) Oh and it runs via Telegram, so I can ask it on the go. WIP of course, prompt engineering needs some work!
00:00
Alex Volkov @ AI Engineer
@altryne
Jan 18, 2024
Just in case you don't want to click over to the other sites, Big Zuck update - Open sourcing will continue - Currently training LLama 3 - AI + Metaverse - Will have 350,000 H100s and ~600 H100 equivalents of compute 🤯 -Ideal AI formfactor is 🕶️
00:00