tweetindex
DE

Andrej Karpathy ✓

@karpathy · Earth · joined 21 Apr 2009

deep learning

4 213 991Followers
1 137Following
10 151Posts total
125MViews on collected posts

Neueste Beiträge

I love this and really hope we can come together as an industry and make it happen. https://t.co/lz2vjjJ5XD
984.5K views · 11.8K likes · 893 reposts · 3 replies Open on X →
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 what Opus 5 would do if I gave it the first paragraph of the Lord of the Rings, a 1M token budget (~$10) and asked for htt
1:32
5.3M views · 28.8K likes · 2.3K reposts · 1.7K replies Open on X →
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, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 4.3M views · 49.9K likes · 4.1K reposts · 2.5K replies Open on X →
@karpathy is there anything special here ? cuz i feel like it has already been a thing for long time, i already have agents i can tag 196.2K views · 628 likes · 2 reposts · 7 replies Open on X →
@karpathy when you guys use this internally do you have channels where only americans are allowed so that it can use mythos 28.7K views · 378 likes · 0 reposts · 7 replies Open on X →
@karpathy We're gonna see token bills like you people have never seen https://t.co/h8bJSGHbef 12.8K views · 226 likes · 1 reposts · 0 replies Open on X →
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 the hood engineering work to make this "just work" (e.g. across tools, integrations, compute environments, memory, 8.3M views · 23.5K likes · 1.9K reposts · 1.4K replies Open on X →
Introducing Claude Tag, a new way for teams to work with Claude. In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work. https://t.co/R2C6A5Kcye
2:24
21M views · 28.4K likes · 2.2K reposts · 1.6K replies Open on X →
In awe of SpaceX and its story - past, present and the future. You can think about it in 10+ different ways and continue re-blowing your mind in circles. Huge congrats to the team! 🚀 1.2M views · 22.3K likes · 1K reposts · 425 replies Open on X →
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 everything by a margin but I'll add that *qualitatively* also, this is a major-version-bump-deserving step change forward 3.1M views · 25.6K likes · 2.4K reposts · 1.3K replies Open on X →
Fable 5 is state-of-the-art on nearly all tested benchmarks, with exceptional performance in software engineering, knowledge work, scientific research, and vision. The longer and more complex the task, the larger Fable 5’s lead over our other models. https://t.co/DxgSu0KUxh
Benchmark table titled Mythos 5 & Fable 5, comparing Claude Mythos 5 and Fable 5 against Claude Mythos Preview, Claude Opus 4.8, GPT 5.5, and Gemini 3.1 Pro.
6.2M views · 15.5K likes · 1.8K reposts · 511 replies Open on X →
@karpathy Same, I have a similar setup. A mix of Obsidian, Cursor (for md), and vibe-coded web terminals as front-end. Since I do a podcast, the number/diversity of research interests is very large. But the knowledge-base approach has been working great. For answers, I often ha 670.3K views · 5.8K likes · 227 reposts · 192 replies Open on X →
@karpathy Can you make a video of how you work with md files, agentic IDEs? Your earlier explanations of using LLMs were really helpful. 450.7K views · 1K likes · 4 reposts · 13 replies Open on X →
@karpathy One thing I've done this year is: - Download all my X data from settings/account - Download all my youtube, gmaps, gmail, google from takout google com - Download all my personal data from Claude, ChatGPT - Export a copy of every AI session on Cursor Claude Code, Codex 79.1K views · 906 likes · 51 reposts · 40 replies Open on X →
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 way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating 21.9M views · 60.9K likes · 7.4K reposts · 2.9K replies Open on X →
@karpathy until that happens, shields up https://t.co/6gboPQgr8y
3.7K views · 119 likes · 6 reposts · 4 replies Open on X →
@karpathy More information here: https://t.co/yEmqG0QbXK Two things that can help: Socket for GitHub (it's free!) to protect your PRs from bad dependencies: https://t.co/D9bsRJj65R Socket Firewall (also free!) to protect your laptop: https://t.co/ebIUWSo0Y9 17.1K views · 99 likes · 7 reposts · 5 replies Open on X →
@karpathy Even if the malicious version was caught early, the real fallout is the trust erosion in the whole npm ecosystem. CI pipelines will now need signed packages or reproducible builds to stay safe. 46.3K views · 276 likes · 5 reposts · 6 replies Open on X →
New supply chain attack this time for npm axios, the most popular HTTP client library with 300M weekly downloads. Scanning my system I found a use imported from googleworkspace/cli from a few days ago when I was experimenting with gmail/gcal cli. The installed version (luckily) 1.6M views · 10.4K likes · 1.1K reposts · 561 replies Open on X →
🚨 CRITICAL: Active supply chain attack on axios -- one of npm's most depended-on packages. The latest axios@1.14.1 now pulls in plain-crypto-js@4.2.1, a package that did not exist before today. This is a live compromise. This is textbook supply chain installer malware. axios 12.5M views · 16K likes · 4K reposts · 536 replies Open on X →
@karpathy Very inspiring as always! We are also open sourcing part of our infra on automated research for Gemini to evolve itself at https://t.co/WH7JBEEm9h More complex than the nanochat setup but closer to SOTA LLM pre/post-training while staying as minimal as possible. More on 105.8K views · 1.4K likes · 144 reposts · 13 replies Open on X →
(I still have the bigger cousin running on prod nanochat, working a bigger model and on 8XH100, which looks like this now. I'll just leave this running for a while...) https://t.co/aWya9hpUMl
459.3K views · 2.1K likes · 58 reposts · 70 replies Open on X →
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 training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the https
11.2M views · 28.2K likes · 3.6K reposts · 1K replies Open on X →
@karpathy "the example proof of the magic of a mind running on 20 watts" 🤯 18K views · 48 likes · 1 reposts · 2 replies Open on X →
@karpathy kimi 2.5 shows how it is still possible for open source to outmatch big labs. 4.4K views · 44 likes · 0 reposts · 2 replies Open on X →
@karpathy i would not have picked up a GPU in my life not for you. such an honor to have your support <3 4.5K views · 71 likes · 0 reposts · 1 replies Open on X →
A conventional narrative you might come across is that AI is too far along for a new, research-focused startup to outcompete and outexecute the incumbents of AI. This is exactly the sentiment I listened to often when OpenAI started ("how could the few of you possibly compete with 1.2M views · 8K likes · 492 reposts · 244 replies Open on X →
Announcing Flapping Airplanes! We’ve raised $180M from GV, Sequoia, and Index to assemble a new guard in AI: one that imagines a world where models can think at human level without ingesting half the internet. https://t.co/7W7WNJ278R
GIF
2.2M views · 3.6K likes · 257 reposts · 337 replies Open on X →
@karpathy I feel this way most weeks tbh. Sometimes I start approaching a problem manually, and have to remind myself “claude can probably do this”. Recently we were debugging a memory leak in Claude Code, and I started approaching it the old fashioned way: connecting a profiler, 1.8M views · 8.2K likes · 543 reposts · 169 replies Open on X →
@karpathy AI didn’t replace programmers. It replaced the programming language. https://t.co/BysT1tPpfc 116.2K views · 1.3K likes · 62 reposts · 32 replies Open on X →

Im Vergleich zu Konten gleicher Größe

3 Beiträge aus den letzten 90 Tagen, verglichen mit der Größenklasse 1M–10M Follower. wird weit über das eigene Publikum hinaus gezeigt, und es reagiert.

Medianaufrufe4 302 798dieses Konto14 648Median für 1M–10M
Reichweite, %102.11%dieses Konto0.74%Median für 1M–10M
Interaktion, %1.32%dieses Konto0.70%Median für 1M–10M
KennzahlDieses KontoMedian für 1M–10MVerhältnis
Medianaufrufe pro Beitrag4 302 79814 648294×
Reichweite (Aufrufe ÷ Follower)102.11%0.74%138×
Interaktionsrate1.32%0.70%1.89×

Weitere Konten dieser Größe →   Mit einem anderen Konto vergleichen →   Wie diese Vergleichswerte entstehen →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

79.1K2 Apr
450.7K
670.3K
6.2M9 Jun
3.1M
1.2M12 Jun
21M23 Jun
8.3M
12.8K
28.7K24 Jun
196.2K
4.3M21 Jul
5.3M2 Aug
984.5K12 Sep

Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

1.27%2 Apr
0.23%
0.93%
0.31%9 Jun
0.95%
1.95%12 Jun
0.16%23 Jun
0.33%
1.78%
1.34%24 Jun
0.33%
1.35%21 Jul
0.64%2 Aug
1.32%12 Sep

Reactions — likes, reposts, replies and quotes — divided by views. Median for 1M–10M accounts is 0.70%.

What the audience does

Likes54.3%355 469 in total
Reposts5.3%34 415 in total
Replies2.4%15 717 in total
Quotes1.9%12 589 in total
Bookmarks36.2%236 898 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

Followers by day

19 Sep22 Sep

Daily snapshots since 19 Sep 2026; the dashed line is the starting count.

Ähnliche Konten