XOutlier XOutlier
@karpathy

Andrej Karpathy

@karpathy

Building @EurekaLabsAI. Previously Director of AI @ Tesla, founding team @ OpenAI.

Open on X ↗
Followers 4.4M Public X audience
Peak Organic Outlier 34.4× Top post vs. median
Highest Post Views 28.1M Single organic post
Outliers Cached 50 781.0K total bookmarks
@karpathy's Top Organic Posts 50 organic posts · Paid boosts excluded
Sorted by Outlier ×
# Date Post ↓
1 May 19 Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here 34.4× 28.1M 149.1K 13.8K
2 Jan 31 I'm being accused of overhyping the [site everyone heard too much about today already]. People's reactions varied very widely, from "how is this interesting at 29.1× 23.8M 21.4K 8.4K
3 Apr 2 LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this wa 26.9× 22.0M 61.0K 108.5K
4 Nov 12 I took delivery of a beautiful new shiny HW4 Tesla Model X today, so I immediately took it out for an FSD test drive, a bit like I used to do almost daily for 5 22× 18.0M 27.4K 4.9K
5 Dec 26 I've never felt this much behind as a programmer. The profession is being dramatically refactored as the bits contributed by the programmer are increasingly spa 20.9× 17.1M 55.3K 32.1K
6 Mar 7 I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM train 13.7× 11.2M 28.2K 38.8K
7 Aug 16 I am (slowly) re-reading the Tolkien legendarium (of which Lord of the Rings is a small part). The whole body of work is so incredible and there's nothing else 11.3× 9.3M 15.4K 4.5K
8 Jun 23 This is a new paradigm for interacting with Claude that is significantly more "inline" with all the other human activity org-wide. Once you do all of the under 10.3× 8.4M 23.5K 13.9K
9 Oct 2 We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks: Writing. Something I've had success with: 9.6× 7.9M 55.6K 81.7K
10 Jan 26 A few random notes from claude coding quite a bit last few weeks. Coding workflow. Given the latest lift in LLM coding capability, like many others I rapidly we 9.6× 7.9M 40.7K 37.0K
11 Apr 4 Wow, this tweet went very viral! I wanted share a possibly slightly improved version of the tweet in an "idea file". The idea of the idea file is that in this e 9.0× 7.3M 26.8K 46.5K
12 Apr 9 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 tr 8.8× 7.2M 22.7K 12.6K
13 Oct 11 Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past ea 8.0× 6.5M 43.2K 29.9K
14 Aug 2 We're starting to leave the territory where you'd test an LLM by e.g. "create an svg of pelican on a bicycle". As one idea to generalize it, I was interested wh 7.6× 6.2M 29.7K 12.1K
15 Oct 13 Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanoc 7.1× 5.8M 24.0K 17.7K
16 Nov 22 As a fun Saturday vibe code project and following up on this tweet earlier, I hacked up an **llm-council** web app. It looks exactly like ChatGPT except each us 6.5× 5.3M 16.9K 13.2K
17 Feb 11 New art project. Train and inference GPT in 243 lines of pure, dependency-free Python. This is the *full* algorithmic content of what is needed. Everything else 6.4× 5.2M 24.8K 28.7K
18 Feb 25 It is hard to communicate how much programming has changed due to AI in the last 2 months: not gradually and over time in the "progress as usual" way, but speci 6.3× 5.2M 36.9K 20.1K
19 Jul 21 One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, bu 5.4× 4.4M 50.2K 15.2K
20 Oct 18 My pleasure to come on Dwarkesh last week, I thought the questions and conversation were really good. I re-watched the pod just now too. First of all, yes I kno 5.0× 4.1M 16.7K 12.7K
21 May 11 This works really well btw, at the end of your query ask your LLM to "structure your response as HTML", then view the generated file in your browser. I've also 5.0× 4.1M 19.4K 21.5K
22 Dec 7 Don't think of LLMs as entities but as simulators. For example, when exploring a topic, don't ask: "What do you think about xyz"? There is no "you". Next time t 4.8× 3.9M 27.4K 17.8K
23 Mar 9 Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes 4.5× 3.7M 19.4K 11.1K
24 Mar 28 - Drafted a blog post - Used an LLM to meticulously improve the argument over 4 hours. - Wow, feeling great, it’s so convincing! - Fun idea let’s ask it to argu 4.3× 3.5M 31.0K 9.3K
25 Feb 20 Bought a new Mac mini to properly tinker with claws over the weekend. The apple store person told me they are selling like hotcakes and everyone is confused :) 4.2× 3.4M 17.4K 9.6K
26 Oct 20 I quite like the new DeepSeek-OCR paper. It's a good OCR model (maybe a bit worse than dots), and yes data collection etc., but anyway it doesn't matter. The mo 4.1× 3.3M 13.2K 6.7K
27 Jun 9 This is a super exciting release - Claude Fable 5 is the same underlying model as Mythos but with added safeguards. The benchmarks are great and it's SOTA on ev 3.9× 3.2M 25.6K 6.3K
28 Nov 23 Gemini Nano Banana Pro can solve exam questions *in* the exam page image. With doodles, diagrams, all that. ChatGPT thinks these solutions are all correct excep 3.7× 3.0M 8.8K 3.1K
29 Dec 28 I was inspired by this so I wanted to see if Claude Code can get into my Lutron home automation system. - it found my Lutron controllers on the local wifi netwo 3.7× 3.0M 25.5K 9.8K
30 Dec 19 2025 LLM Year in Review 2025 has been a strong and eventful year of progress in LLMs. The following is a list of personally notable and mildly surprising "parad 3.7× 3.0M 15.5K 17.5K
31 Nov 18 I’m starting to get into a habit of reading everything (blogs, articles, book chapters,…) with LLMs. Usually pass 1 is manual, then pass 2 “explain/summarize”, 3.6× 2.9M 13.3K 8.1K
32 Mar 25 One common issue with personalization in all LLMs is how distracting memory seems to be for the models. A single question from 2 months ago about some topic can 3.4× 2.8M 20.9K 3.3K
33 Apr 30 This is the the quote I've been citing a lot recently. 3.4× 2.8M 46.7K 11.4K
34 Jun 6 Making slides manually feels especially painful now that you know Cursor for slides should exist but doesn’t. 3.3× 2.7M 12.1K 3.2K
35 Jul 23 Love this! Supercharger, diner, … but really a kind of exhibit for the future. Plotting a road trip SF -> LA to charge Shadowfax 3.2× 2.6M 13.4K 550
36 Sep 5 I think congrats again to OpenAI for cooking with GPT-5 Pro. This is the third time I've struggled on something complex/gnarly for an hour on and off with CC, t 3.2× 2.6M 12.4K 2.6K
37 Nov 21 Something I think people continue to have poor intuition for: The space of intelligences is large and animal intelligence (the only kind we've ever known) is on 3.2× 2.6M 11.3K 8.6K
38 Jun 18 Part 2 of this mystery. Spotted on reddit. In my test not 100% reproducible but still quite reproducible. 🤔 3.2× 2.6M 9.1K 3.4K
39 Feb 25 With the coming tsunami of demand for tokens, there are significant opportunities to orchestrate the underlying memory+compute *just right* for LLMs. The fundam 3.1× 2.6M 7.3K 4.6K
40 Nov 24 A number of people are talking about implications of AI to schools. I spoke about some of my thoughts to a school board earlier, some highlights: 1. You will ne 3.1× 2.5M 16.4K 8.5K
41 Mar 11 Expectation: the age of the IDE is over Reality: we’re going to need a bigger IDE (imo). It just looks very different because humans now move upwards and progra 3.1× 2.5M 10.4K 4.7K
42 Mar 26 When I built menugen ~1 year ago, I observed that the hardest part by far was not the code itself, it was the plethora of services you have to assemble like IKE 3.1× 2.5M 6.4K 4.0K
43 Jun 25 +1 for "context engineering" over "prompt engineering". People associate prompts with short task descriptions you'd give an LLM in your day-to-day use. When in 2.9× 2.4M 14.3K 9.2K
44 Sep 25 "AI isn't replacing radiologists" good article Expectation: rapid progress in image recognition AI will delete radiology jobs (e.g. as famously predicted by Geo 2.9× 2.3M 8.5K 6.0K
45 Feb 24 CLIs are super exciting precisely because they are a "legacy" technology, which means AI agents can natively and easily use them, combine them, interact with th 2.7× 2.2M 11.7K 9.4K
46 Nov 16 Sharing an interesting recent conversation on AI's impact on the economy. AI has been compared to various historical precedents: electricity, industrial revolut 2.6× 2.1M 12.3K 8.4K
47 Aug 18 I get ~10 spam calls per day (various automated voicemails, "loan pre-approval" etc) and ~5 spam messages per day (usually phishing). - I have AT&T Active Armor 2.5× 2.1M 16.5K 2.1K
48 Oct 1 Finally had a chance to listen through this pod with Sutton, which was interesting and amusing. As background, Sutton's "The Bitter Lesson" has become a bit of 2.4× 2.0M 9.5K 8.6K
49 Feb 19 Very interested in what the coming era of highly bespoke software might look like. Example from this morning - I've become a bit loosy goosy with my cardio rece 2.4× 2.0M 12.0K 7.8K
50 Oct 16 TV in the 90s: you turn it on, you watch. TV 2025: - turn on, wait for it to load - popup: TV wants to update, 1.5GB. No. - scroll sideways, find prime video ap 2.1× 1.7M 22.0K 1.3K
Compare other creators:
@levelsio@paulg@naval@marclou@illyism