# The Best AI Twitter/X Accounts to Follow > The best AI Twitter/X accounts to follow in 2026: researchers, builders, lab leaders, and signal-not-hype feeds across LLMs, ML, and applied AI. - Source: https://zarifautomates.com/blog/best-ai-twitter-x-accounts-to-follow - Published: 2026-07-04 - Updated: 2026-07-04 - Pillar: AI News & Trends - Tags: ai-twitter, ai-influencers, ai-news, machine-learning, social-media - Author: Zarif --- X is still where AI breaks first. New papers, model releases, leaks, benchmark debates — it all hits Twitter hours before it hits any newsletter or news site. The question is not whether to use X for AI news. The question is who to follow. "AI Twitter" or "AI X" refers to the loosely connected community of researchers, lab founders, engineers, journalists, and builders who post about artificial intelligence on the platform formerly known as Twitter. It is the de facto real-time signal layer for the AI industry — paper releases, model launches, benchmark debates, and policy fights play out there before they appear anywhere else. - Andrej Karpathy (1.4M followers) is still the highest signal-to-noise account on AI X — when he posts, it is worth reading - For research depth: Yann LeCun, Demis Hassabis, Jeff Dean, Lilian Weng, and Sebastian Raschka. For builders: Andrew Ng, Santiago Valdarrama, Simon Willison, Swyx - For lab announcements: follow @OpenAI, @AnthropicAI, @GoogleDeepMind, @MetaAI, @Mistralai directly. Lab CEOs (Sam Altman, Dario Amodei, Demis Hassabis) post strategic context - For news curation: @rowancheung, @deedydas, @bindureddy, @emollick. They aggregate the firehose into something readable - Aim for 30 to 50 follows max. More than that and the timeline becomes noise ## How to Use This List The biggest mistake newcomers to AI X make is following too many accounts. The platform's algorithm rewards engagement, and engagement on AI X often means hot takes, not insight. If you follow 200 AI accounts, your feed becomes a stream of dunks and threadboi rage-bait. Better approach: follow 30 to 50 accounts across four categories — researchers, builders, lab voices, and curators — and use Lists to separate them. Mute the engagement farmers ruthlessly. The accounts below are organized by what role they play in your feed. ## The Researchers (Signal, Not Noise) These are the people whose posts make you smarter when you read them. **Andrej Karpathy (@karpathy)** — Former Director of AI at Tesla, founding member of OpenAI, now running Eureka Labs. About 1.4M followers. Karpathy posts intuition, learning advice, and perspective on where the field is going. His tutorials and "build a GPT from scratch" videos shaped how a generation of engineers thinks about transformers. When Karpathy posts, you read. **Yann LeCun (@ylecun)** — Chief AI Scientist at Meta, ACM Turing Award laureate, NYU professor. About 940K followers. LeCun is famously skeptical of LLM-only paths to AGI and pushes for world models, JEPA, and self-supervised learning. He argues with everyone, often productively. Follow for the alternative perspective on scaling. **Demis Hassabis (@demishassabis)** — Co-founder and CEO of Google DeepMind, 2024 Nobel Prize laureate. Posts measured updates on DeepMind research, AlphaFold, Gemini, and AI for science. Lower volume than most but consistently substantive. **Jeff Dean (@JeffDean)** — Chief Scientist of Google DeepMind and Google Research, Gemini lead. The legendary architect behind a huge chunk of modern Google infrastructure. Posts on systems, scaling, and Gemini. **Lilian Weng (@lilianweng)** — Research Scientist at DeepMind, formerly Head of Safety at OpenAI. Her blog posts on RLHF, agents, and prompt engineering are widely cited. Twitter is where she previews and discusses them. **Sebastian Raschka (@rasbt)** — Author of *Build a Large Language Model From Scratch* and *Machine Learning with PyTorch*. About 200K followers. Best account on X for hands-on implementation breakdowns of new architectures. If you want to actually understand how an LLM is trained at the code level, follow Raschka. **Fei-Fei Li (@drfeifei)** — Stanford professor, Co-Director of Stanford HAI, former Chief Scientist of AI/ML at Google Cloud. About 516K followers. Posts on AI policy, spatial intelligence, and the human side of AI development. **François Chollet (@fchollet)** — Creator of Keras, formerly at Google. Posts skeptical, well-argued takes on AGI, benchmarks (he created ARC-AGI), and why current LLMs are not the whole story. **Jim Fan (@DrJimFan)** — Senior Research Manager at Nvidia, Embodied AI lead. Threads on robotics, reinforcement learning, and foundation models that go viral and deserve to. ## The Builders (Practical, Hands-On) These accounts are less about pushing research and more about shipping. **Andrew Ng (@AndrewYNg)** — Founder of DeepLearning.AI, Coursera, Landing AI. About 1.1M followers. Ng's posts are reliable, calm signal in an otherwise hot ecosystem. Strong on data-centric AI, applied ML, and educational content. **Santiago Valdarrama (@svpino)** — Computer scientist, formerly Principal ML Engineer at Microsoft. About 380K followers. Teaches ML through clear, practical Twitter posts. Particularly strong on system design and the gap between toy ML and production ML. **Simon Willison (@simonw)** — Co-creator of Django, builds Datasette and llm CLI. Best running commentary on what new LLMs can and cannot actually do. His blog (simonwillison.net) is mandatory reading; his X is the live version. **Shawn @swyx (@swyx)** — Curator of Latent Space podcast and AI Engineer Summit. Posts on the AI engineering stack — agents, RAG, evals, embeddings. Probably the best single account for understanding "how do builders actually use AI today." **Riley Goodside (@goodside)** — Staff Prompt Engineer at Scale AI. The person who effectively invented modern prompt engineering. Follow for clever, weird LLM prompting demonstrations. **Mckay Wrigley (@mckaywrigley)** — Builder of Takeoff AI School and Chatbot UI. Posts daily on how he uses AI to ship faster. Practical, less academic. **Logan Kilpatrick (@OfficialLoganK)** — Lead Product at Google AI Studio, formerly OpenAI Developer Relations. The most useful single account for staying on top of API changes, model releases, and developer-facing AI updates. **Aravind Srinivas (@AravSrinivas)** — CEO of Perplexity. Posts product strategy, AI search, and frequent reflections on competing with Google. ## The Lab Voices (Strategic Context) These accounts shape the narrative of what AI labs are doing and why. **Sam Altman (@sama)** — CEO of OpenAI. Posts shape narratives, occasionally moves markets, and offers strategic context on where OpenAI is pointed. Read for signal even when (especially when) the rhetoric is grand. **Dario Amodei (@DarioAmodei)** — CEO of Anthropic. Lower volume than Altman, denser content. His "Machines of Loving Grace" essay shaped industry thinking; his Twitter is where he previews framing. **Mira Murati (@miramurati)** — Founder of Thinking Machines Lab, formerly CTO of OpenAI. Posts strategic and product-level perspectives. **Greg Brockman (@gdb)** — President of OpenAI, co-founder. Often previews demos and shares OpenAI culture context. **Mustafa Suleyman (@mustafasuleyman)** — CEO of Microsoft AI, co-founder of DeepMind and Inflection. Posts on AI policy, the existential risk debate, and product-level Microsoft AI direction. **Emad Mostaque (@EMostaque)** — Founder of Stability AI (former), now Schelling AI. Posts on open-source AI, and frequent industry commentary. **Lab accounts to follow directly**: @OpenAI, @AnthropicAI, @GoogleDeepMind, @MetaAI, @Mistralai, @huggingface, @nvidia, @cohere, @perplexity_ai, @togetherai. These are where official launches drop first. Use X's "Lists" feature aggressively. Make a private list called "AI Researchers" with 15 to 20 accounts and another called "AI Builders" with 15 to 20. Browse the lists instead of your home timeline when you want signal. The home algorithm will serve you 5x more outrage than the lists will, by design. ## The News Curators (Aggregators You Can Trust) The firehose of AI X is unmanageable. These accounts curate it. **Rowan Cheung (@rowancheung)** — Founder of The Rundown AI, the world's largest independent AI newsletter at 1.75M+ readers. His X feed previews what The Rundown will cover and surfaces under-the-radar tools. **Deedy Das (@deedydas)** — Investor at Menlo Ventures. Posts the best one-tweet summaries of new model releases and benchmarks. Reliable filter for "is this actually new or is it noise." **Bindu Reddy (@bindureddy)** — CEO of Abacus AI. Posts heated takes on AI capabilities and progress. Not always right, but always interesting. **Ethan Mollick (@emollick)** — Wharton professor, author of *Co-Intelligence*. Best account for "what does AI actually do for knowledge workers right now." Empirical, practical, calmly written. **Linus Lee (@thesephist)** — Researcher at Notion. Threads on LLM internals, embeddings, and AI-native software design. Underrated. **Nathan Labenz (@labenz)** — Cognitive Revolution podcast host, AI red-teamer. Posts thoughtful threads on capabilities, alignment, and current model evaluations. **Dwarkesh Patel (@dwarkesh_sp)** — Host of Dwarkesh Podcast. The single best long-form interview show in AI. His X is where he discusses guests and arguments. **Nathan Lambert (@natolambert)** — Researcher at Allen Institute for AI, runs the Interconnects newsletter. Best curator on RLHF, post-training, and the open-weight ecosystem. **Jack Clark (@jackclarkSF)** — Co-founder of Anthropic, writes the Import AI newsletter. Reliable, measured, government and policy lens. ## Research-Focused Accounts and Aggregators For people who want primary research signal: **@DAIR_AI** — Curates AI papers with thread explainers. Reliable feed for staying current on ML research. **@arxiv_org and @alphaxiv** — Direct from arXiv, with AlphaXiv adding social discovery on papers. **@_akhaliq** — Aggregates the best new papers daily. Mostly retweets, but excellent filter. **@hardmaru (David Ha)** — Researcher, formerly at Google Brain and Stability AI. Eclectic, thoughtful, mixes research and aesthetics. **@cwolferesearch (Cameron Wolfe)** — Senior Research Scientist at Netflix. Writes excellent paper breakdowns and the Deep (Learning) Focus newsletter. **@vikhyatk** — Builder of Moondream (small VLM). Underrated voice on practical small-model engineering. ## The Comparison Table
Category Account What You Get Volume
Research @karpathy Deep intuition, learning advice, frontier perspective Low, very high signal
Research @ylecun Skeptical takes on scaling, world models, debates High
Research @demishassabis DeepMind direction, science applications Low
Research @rasbt Hands-on LLM implementation breakdowns Medium
Builders @AndrewYNg Applied AI, education, calm signal Medium
Builders @simonw What LLMs actually do, live testing High
Builders @swyx AI engineering stack, agents, evals High
Lab CEOs @sama OpenAI strategy, narrative-shaping Medium
Lab CEOs @DarioAmodei Anthropic strategy, alignment framing Low
Curators @emollick Empirical "what does AI actually do" posts Medium
Curators @rowancheung Tools, news aggregation, daily highlights High
Curators @deedydas Sharp model-release summaries Medium
## Accounts to Be Cautious With Some of the highest-engagement AI accounts on X are net-negative for understanding the field. **The hype-bots**: accounts that post "10 AI tools that will change your life" threads daily. They are usually drop-shipping affiliate links, and the tools rotate based on commission, not quality. **The doomer-grifters and accelerationist-grifters**: there are people on both sides of the AI safety debate who have learned the engagement loop and now produce identical hot takes weekly. Some are sincere; many are not. If an account posts 20+ times a day about AGI timelines, mute it. **Anonymous "leaks" accounts**: occasionally right, frequently wrong, always confidently posted. Check leaks against actual primary sources before sharing. **The "vibes" researchers**: accounts that post feelings about model capabilities without ever showing prompts, outputs, or evals. The good news is the actual researchers (Karpathy, Mollick, Willison) all show their work — copy that pattern. The general rule: if an account is not posting code, papers, screenshots, or specific prompts, you are reading marketing. ## How to Build Your Following Stack A practical recipe: 1. **Start with 5 anchors**: Karpathy, Ng, Mollick, Willison, swyx. These five give you a baseline of research, applied, empirical, hands-on, and engineering perspectives. 2. **Add 10 lab voices**: OpenAI, Anthropic, DeepMind, Meta AI, Mistral, plus their CEOs and a few key researchers. Now you catch every major release. 3. **Add 10 specialist researchers**: pick the area you care about (RLHF, vision, robotics, agents) and find the 10 best researchers in it. Sebastian Raschka, Lilian Weng, Jim Fan, Cameron Wolfe, Nathan Lambert is a good starter set. 4. **Add 5 curators**: deedydas, rowancheung, akhaliq, DAIR_AI, alphaxiv. These give you the firehose pre-filtered. 5. **Mute aggressively**: anyone who appears in your feed for outrage value gets muted. Lists and mute lists are the actual product. Total: about 30 accounts. You will see the signal without drowning. ## Frequently Asked Questions ## Related Guides - [The Best AI Blogs and Websites for News](/blog/best-ai-blogs-and-websites-for-news) - [The Best AI Discord Servers for Networking](/blog/best-ai-discord-servers-for-networking) - [The Best AI Subreddits for Discussion](/blog/best-ai-subreddits-for-discussion) **Is X still the right platform for AI news in 2026?** For real-time signal, yes — there is no replacement for AI X. Lab releases, paper threads, and benchmark debates still hit X first. The platform itself has problems (algorithmic chaos, hostility, bot accounts) but the AI community has not migrated en masse to Mastodon, Bluesky, or Threads. If you are serious about AI, you need an X account, even if you only read. Use Lists to make it tolerable. **How many AI accounts should I actually follow?** 30 to 50 is the sweet spot. Below 30 and you miss obvious news. Above 50 and your feed becomes noise — you scroll past important posts because there are too many. The discipline is unfollowing accounts that no longer earn the timeline slot. Audit every 6 months and prune. **Should I follow lab accounts (@OpenAI, @AnthropicAI) directly?** Yes, follow the official lab accounts for launches and announcements. But do not rely on them for analysis — they are PR feeds. Pair them with researcher accounts who provide context. The pattern: @OpenAI announces a launch, @karpathy or @swyx explains what it actually means, @simonw tests it live and shows the gotchas. All three together is the full picture. **Are AI influencers on X reliable?** Some are, most are not. Reliable signals: they show their work (code, prompts, screenshots), they are calibrated about uncertainty, and they correct themselves when wrong. Unreliable signals: they sell a course, they post 20 times a day, they treat every release as either world-ending or world-saving. The accounts in this article have been filtered for the reliable signals. Trust but verify — for any specific claim, check the primary source. **What about AI YouTubers and podcasters who are also on X?** Several of them are excellent X follows independent of their other content — Dwarkesh Patel, Nathan Labenz, Lex Fridman are good examples. The X feed often previews podcast guests and offers live reactions to news. If you already follow their long-form content, the X account is a useful accompaniment rather than a substitute. ## The Bottom Line AI X is the most valuable, most exhausting, most signal-rich, most noise-saturated information source in the field. Used well — with Lists, mute filters, and a curated 30 to 50 follows — it is a competitive advantage. Used poorly — with 500 follows on the home timeline — it is a treadmill of outrage. Pick 5 anchor accounts from this list today. Build out from there over the next month. Audit quarterly. Mute liberally. The signal exists. You just have to filter for it. --- **More AI news and trend reading:** See [Best AI Newsletters to Subscribe To](/blog/best-ai-newsletters-to-subscribe-to), [Best AI Podcasts for Staying Informed](/blog/best-ai-podcasts-for-staying-informed), and [Best AI YouTube Channels for Education](/blog/best-ai-youtube-channels-for-education).