Here's my conversation all about @FFmpeg, the legendary open-source software powering most video on the Internet. In the episode, I talk with Jean-Baptiste Kempf and Kieran Kunhya. JB is lead developer of VLC and Kieran is FFmpeg contributor, codec engineer, and the person behind the now-infamous @FFmpeg account on X.
VLC (@videolan), by the way, is also a legendary piece of open-source software: it's a video player that can open basically anything & has been downloaded over 6 billion times.
I think both FFmpeg and VLC are two of the most important and impactful software systems ever created, both open source, and both created & maintained by volunteers: brilliant engineers from all walks of life.
Thank you to everyone who contributed to FFmpeg and VLC, and in general to all engineers giving their heart & soul to building systems used by millions (or billions) of people, and often doing so not for money, status, or fame, but purely for the love of building great software and doing good for the world.
Thank you to the builders! 🙏❤️
Shoutouts in this chat to @ID_AA_Carmack@karpathy@elonmusk@TimSweeneyEpic and everyone who is a contributor & fan of open source!
It's here on X in full and is up everywhere else (see comment).
Timestamps:
0:00 - Episode highlight
2:17 - Introduction
5:35 - Weirdest things VLC opens
9:59 - How video playback works
19:20 - Video codecs and containers
30:07 - FFmpeg explained
51:07 - Linus Torvalds
55:46 - Turning down millions to keep VLC ad-free
1:10:04 - FFmpeg & Google drama
1:29:18 - FFmpeg developers
1:35:55 - VLC and FFmpeg
1:40:29 - History of FFmpeg
1:43:46 - Reverse engineering codecs
1:57:01 - FFmpeg testing
2:01:08 - Assembly code (handwritten)
2:25:26 - Rust programming language
2:34:42 - FFmpeg and Libav fork
2:43:04 - Open source burnout
2:50:51 - x264 and internet video
3:04:07 - Video compression basics
3:11:04 - CIA and fake VLC
3:21:39 - Ultra low latency streaming
3:39:07 - AV2 codec and video patents
3:48:59 - VLC backdoors
3:59:14 - Video archiving
4:05:51 - Future of FFmpeg and VLC
Tech teachers start teaching now how IShowSpeed streaming on places where even startlinks not working yet how his production is able to do all that which big productions not able to do yet 🔥🔥
I had an accident last week and took a shot to the eye. My old white optometrist got me in within 2 hours and told me I had a torn retina and I likely needed surgery.
And told me not to wait. It could worsen and threaten my vision.
The next morning, I got laser surgery done by an Indian doctor, and it was successful, and I'm doing fine.
Every day I waited risked serious harm.
A couple of months ago, DD News (yes, the sarkari one) reached out for my thoughts on AI.
Didn’t expect this to turn into a genuinely world-class documentary.
Link in first reply, worth your time. @DDNewslive@DDIndialive
Some Cinephiles asked me : “What’s the real role of film criticism—and who are its greatest voices?”
Start here:
• André Bazin’s groundbreaking What is Cinema? (the bible of realism)
• Everything by his brilliant pupil François Truffaut
• Godard on Godard
• Pauline Kael’s I Lost It at the Movies
• The essential writings of the great Bengali critic Chidananda Dasgupta
These will sharpen how you see, feel, and think about cinema forever. 📽️
🚨 BREAKING: Someone just open-sourced software that sees you through walls using only WIFI signals.
it’s called WiFi-DensePose. It maps your exact body pose in real-time. no cameras. no sensors. just your living room router.
100% Open Source.
It does appear that @Alibaba_Qwen’s Qwen3.5-27B has already achieved what DeepSeek V4 is rumored to be, a low cost near SOTA level LLM to shake up BigAI. A full unquantized version can run on a beefy home PC w/64GB of RAM at no API costs.
I downloaded and ran Qwen3.5-122B today in a 4-bit quant variant, locally on a GPU. Outstanding reasoning capabilities and really astonishing throughput. Getting about 22 t/s on text generation which is extraordinary for a 122B model on local workstation hardware. Coding tasks so far look strong. Clear reasoning capabilities in place, code overview creates strategies for achieving output goals. It finds errors and fixes them. It suggests improvements. This doesn't even beat Opus 4.6 on SWE bench but it's seriously good and LOCAL. I connected Claude Code to it so now I'm running Claude Code for free locally (and will connect with Opus if the local solution balks at anything).
You can now vibecode your own WisprFlow or Monologue alternative that runs completely locally on Apple Silicon using MLX-Audio-Swift 🔥
Check out this live transcription of @dwarkesh_sp interview with @karpathy using Qwen3-ASR-0.6B quantized to 4bit on a M3 Max.
It also runs in realtime on a iPhone 15 Pro and iPad Pro M1.
No cloud. No API keys.
https://t.co/AXJvHw0DY6
🚨 BREAKING: Someone just made 70B parameter models run on a single 4GB GPU.
It's called AirLLM. No quantization. No distillation. No pruning. Just raw 70B inference on hardware that costs less than a dinner.
You can even run Llama 3.1 405B on 8GB VRAM.
Here's how it works:
→ Decomposes the model layer-by-layer
→ Loads only one layer into GPU memory at a time
→ Runs inference, moves to the next layer
→ Prefetches the next layer while computing the current one
→ Supports 4-bit and 8-bit compression for 3x speed boost
No cloud API. No $10K GPU. Just pip install airllm and go.
Here's the wildest part:
It supports almost every major model — Llama, Qwen, Mistral, ChatGLM, Baichuan, InternLM — and it auto-detects the model type. One line of code to load. One line to generate.
Works on Linux, macOS (Apple Silicon), and even Google Colab free tier.
Your old gaming laptop can now run the same models that needed an A100.
100% Open Source. Apache 2.0 License.
Demis Hassabis’s “Einstein test” for defining AGI:
Train a model on all human knowledge but cut it off at 1911, then see if it can independently discover general relativity (as Einstein did by 1915);
if yes, it’s AGI.
https://t.co/IzBDZveeJf is a math website for learning functions, geometry, algebra, CALCULUS, statistics, and 3D math, like https://t.co/5JPi2FcjRU.
It's a great site for kids to learn calculus.
I have a customer with a ton of PDFs they want an LLM on top of, but we're hitting context window limits
Is there a high-level API that lets me upload a bunch of PDFs, and then provides a "tool" that I can give to an LLM?
Meghna Gulzar, the darling of the left loony cabal, was a minor when she rammed her car injuring a person. She was only 14! That person survived and could narrate his story.
Read this decade old piece…
https://t.co/TwAP3vYLcI