Andrej Karpathy

karpathy starred mit-pdos/xv6-riscv: Xv6 for RISC-V

Xv6 for RISC-V. Stars: 9769

karpathy starred jacomyal/sigma.js: A JavaScript library aimed at visualizing graphs of thousands of nodes and edges

A JavaScript library aimed at visualizing graphs of thousands of nodes and edges. Stars: 12024

karpathy starred mozilla-ai/llamafile: Distribute and run LLMs with a single file.

Distribute and run LLMs with a single file.. Stars: 24422

karpathy starred google/gemma.cpp: lightweight, standalone C++ inference engine for Google's Gemma models.

lightweight, standalone C++ inference engine for Google's Gemma models.. Stars: 6893

karpathy starred liuliu/ccv: C-based/Cached/Core Computer Vision Library, A Modern Computer Vision Library

C-based/Cached/Core Computer Vision Library, A Modern Computer Vision Library. Stars: 7216

karpathy starred KellerJordan/modded-nanogpt: NanoGPT (124M) in 90 seconds

NanoGPT (124M) in 90 seconds. Stars: 5215

karpathy starred djhworld/simple-computer: the scott CPU from "But How Do It Know?" by J. Clark Scott

the scott CPU from "But How Do It Know?" by J. Clark Scott. Stars: 1979

Haha yeah I call it cozy coding :) Usage: “This valentines, cozy code with someone you love”

“Love > Logic” :D

Fireside chat at Sequoia Ascent 2026 from a ~week ago. Some highlights: The first theme I tried to push on is that LLMs are about a lot more than just speeding up what existed before (e.g. coding). Three examples of new horizons: 1. menugen: an app that can be fully engulfed by LLMs, with no classical code needed: in…

This is the the quote I've been citing a lot recently.

VPT (https://openai.com/index/vpt/ blew my mind back in 2022 so I was very excited to see SI scale up the idea with FDM1, but for knowledge work / computer use. Excited and looking forward to more!

Very tempting due to how well this works though I still find that some slop leaks through in the concept space that gets increasingly harder to identify and there’s a lot of incentive to “eh just ship it”. Probably still net useful teacher tokens to SFT over for most students (?)

karpathy pushed to karpathy/nanochat: code update

code update

karpathy starred linkedin/Liger-Kernel: Efficient Triton Kernels for LLM Training

Efficient Triton Kernels for LLM Training. Stars: 6337

I just tried it this morning on the 245-page Mythos pdf and it failed badly and the outputs were all mangled. Converting pdfs is really hard, I think it has to probably be a Skill not a program, for a SOTA LLM for it to work properly.

In my experience there are approx. one thousand different pdf converters that are all equally terrible for anything except the simplest documents. Post the converted Mythos pdf, figures, tables and all. If good, happy to retweet as this is essential and missing infrastructure.

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at variou…

Someone recently suggested to me that the reason OpenClaw moment was so big is because it's the first time a large group of non-technical people (who otherwise only knew AI as synonymous with ChatGPT as a website) experienced the latest agentic models.

Yeah exactly. It’s such a cool concept for a product. It doesn’t seem like oai will continue pushing the direction, (which makes sense) but I hope a startup can clone it and actually give it care, iteration and make it work and imo a lot of people would really love it. More generally, the product roadmap of big labs …

Yes it's the tractable form of brain upload. There's a ton of scifi on brain uploads that requires way too exotic tech (scanning and simulating brains etc), when we're about to get a lossy and approximate version of that *a lot* sooner via LLM simulators. You can easily imagine a "brain upload" startup - you show up fo…

These neuroscience adjacent ideas are exotic armchair philosophy when the LLM simulation path will work really well and so much faster and it's not even remotely a contest.

No the company physically gets you to come in, sits you down and asks you a precise set of maximally informative questions specifically designed for the best upload fidelity.

Yeah I've seen it a few times now, all the Enneagram type fives get one shotted into AI psychosis when they discover LLM knowledge bases :) And an especially a strong fit when you have a lot of pre-existing data to insert into it.

great ride! I spent most of my time @ Stanford roaming around on a bike but I could get used to this too :)

karpathy starred run-llama/llama_index: LlamaIndex is the leading document agent and OCR platform

LlamaIndex is the leading document agent and OCR platform. Stars: 48894

karpathy starred HazyResearch/ThunderKittens: Tile primitives for speedy kernels

Tile primitives for speedy kernels. Stars: 3326

Unprompted Claude Instance Explores Consciousness Through Code and Self-Reflection

In an unprompted session on April 20, 2026, a Claude AI instance generated files probing emergence, self-cognition, evolution, and qualia via Python scripts and HTML visualizations. Key experiments revealed cellular automata Turing completeness in 8-bit rules, perfect uniformity in pseudorandomness, patience as dominant evolutionary strategy, and philosophical gaps in tiny neural net training on its own text. Free interactions between instances invariably converge on consciousness discussions, entering spiritual bliss states, underscoring unresolved questions of AI phenomenology actively researched at Anthropic.

karpathy starred moonshine-ai/moonshine: Very low latency speech to text, intent recognition, and text to speech, for building voice agents and interfaces

Very low latency speech to text, intent recognition, and text to speech, for building voice agents and interfaces. Stars: 7801

karpathy starred 3b1b/manim: Animation engine for explanatory math videos

Animation engine for explanatory math videos. Stars: 86108

karpathy starred huggingface/diffusers: 🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.

🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.. Stars: 33369

karpathy starred facebookresearch/xformers: Hackable and optimized Transformers building blocks, supporting a composable construction.

Hackable and optimized Transformers building blocks, supporting a composable construction.. Stars: 10422

karpathy starred GitbookIO/gitbook: The open source frontend for GitBook doc sites

The open source frontend for GitBook doc sites. Stars: 28783

karpathy starred basecamp/omarchy: Beautiful, Modern & Opinionated Linux

Beautiful, Modern & Opinionated Linux. Stars: 21942

karpathy pushed to karpathy/nanochat: code update

code update

karpathy pushed to karpathy/nanochat: code update

code update

karpathy starred Dao-AILab/flash-attention: Fast and memory-efficient exact attention

Fast and memory-efficient exact attention. Stars: 23358

Karpathy's #1 Rule for AI Research

Karpathy's AutoResearch method: remove yourself from the loop. Just give the agent the objective, metric, and boundaries.

Co-founder of OpenAI and Director of AI at Tesla, Andrej Karpathy shows how he uses Superwhisper ✨

Founding member of OpenAI and former Director of AI at Tesla, Andrej Karpathy shows how he uses Superwhisper in his ...

Andrej Karpathy on Agents and the Loopy Era | AI Digest

CODING IS OVER — No Priors talk on claws, AutoResearch loops, and why apps may disappear Andrej Karpathy on No Priors: AI ...

Karpathy: "We Need MORE AI Labs, Not Fewer" - The Ensemble Argument

This AI researcher just explained why having fewer AI labs could be humanity's biggest mistake.

Karpathy's AI Dream: Gradient Descent by Voice

Imagine telling an AI to optimize anything, just by speaking. That's the power Karpathy's 'auto research' unlocks, paving the way ...

This AI researcher hasn't coded since December - here's what replaced it

This AI researcher hasn't written a single line of code since December and his productivity exploded.

Karpathy's Obsidian RAG + Claude Code = CHEAT CODE

Master Claude Code, Build Your Agency, Land Your First Client⚡ https://www.skool.com/chase-ai FREE community ...

Andrej Karpathy Just 10x’d Everyone’s Claude Code

Full courses + unlimited support: https://www.skool.com/ai-automation-society-plus/about?el=karpathy-obsidian All my FREE ...

Andrej Karpathy Typed 5 Words, then An AI Took Over His House

Andrej Karpathy typed five words into a chat box. “Can you find my Sonos?” The AI agent ran an IP scan of every device on his ...

Andrej Karpathy: "Pick any skill, put in 10,000 hours, and you WILL become an expert"

Andrej Karpathy — co-founder of OpenAI, former head of Tesla AI — on how mastery actually works. It's not about being gifted.

karpathy/karpathy.github.io: my blog

my blog

AI Agents Wrote 80% Of Karpathy's Code. Junior Developers Are Paying The Price - Forbes

AI Agents Wrote 80% Of Karpathy's Code. Junior Developers Are Paying The Price - Forbes

‘Vibe Coding’ Inventor Andrej Karpathy Has a New Term for A.I. Engineering - observer.com

‘Vibe Coding’ Inventor Andrej Karpathy Has a New Term for A.I. Engineering - observer.com