# Best AI Twitter (X) Accounts to Follow in 2026 > A curated list of the best AI Twitter and X accounts for research, engineering, product news, policy, and practical analysis in 2026. - Source: https://www.zarifautomates.com/blog/best-ai-twitter-x-accounts-to-follow - Published: 2026-07-04 - Updated: 2026-08-12 - Pillar: AI News & Trends - Tags: ai-twitter, ai-influencers, ai-news, machine-learning, social-media - Author: Zarif --- AI Twitter—now AI X—is still useful because researchers, builders, and labs often post work at the moment it becomes public. It is also full of recycled demos and confident claims without evidence. The solution is not a bigger following list. It is a balanced one: primary lab accounts for announcements, technical people who show their work, and a few analysts who add context. AI Twitter or AI X is the network of researchers, engineers, founders, educators, policymakers, and official organizations discussing artificial intelligence on X. The best accounts link to papers, code, documentation, evaluations, or firsthand experiments instead of merely repeating news. - Start with **Andrej Karpathy, Simon Willison, Ethan Mollick, Sebastian Raschka, and Chip Huyen** - Follow official lab accounts for primary announcements, then use independent analysts for interpretation - Separate research, engineering, products, and policy into private X Lists - Ignore follower counts; evaluate whether an account links evidence and corrects mistakes - Twenty deliberate follows are more useful than a feed of hundreds of AI accounts ## Best AI X Accounts at a Glance
Account Best for Typical value
@karpathy LLM intuition and education Original explanations and experiments
@simonw Hands-on model testing Fast, documented product analysis
@emollick AI at work and education Research-informed practical guidance
@rasbt How language models work Code-level technical explanations
@chipro Production AI systems Engineering frameworks and tradeoffs
@AndrewYNg Applied AI and learning Accessible industry and education perspective
@swyx AI engineering Tools, events, and builder trends
@natolambert Open models and post-training Research commentary and synthesis
@DrJimFan Robotics and embodied AI Research explanations and demos
@fchollet Reasoning and AI evaluation Critical benchmark perspective
## The Five Best Starting Accounts ### 1. Andrej Karpathy — @karpathy [Follow @karpathy](https://x.com/karpathy). Karpathy is the strongest general-purpose starting point for people who want intuition about neural networks, language models, tokenization, training, and AI-assisted coding. His most useful posts usually connect to a longer explanation, repository, talk, or experiment. **Why follow:** he turns technical mechanisms into mental models without reducing everything to a hot take. ### 2. Simon Willison — @simonw [Follow @simonw](https://x.com/simonw). Willison tests new language-model products and APIs quickly, records prompts and outputs, and links back to detailed notes on his website. This makes his feed especially useful during launches, when marketing summaries spread faster than careful testing. **Why follow:** he shows what happened, preserves evidence, and distinguishes observed behavior from speculation. ### 3. Ethan Mollick — @emollick [Follow @emollick](https://x.com/emollick). Mollick focuses on how people use AI in work and education. His posts are useful for managers, teachers, consultants, and knowledge workers who need practical implications rather than implementation details. **Why follow:** he frequently connects recommendations to research and real classroom or workplace use. ### 4. Sebastian Raschka — @rasbt [Follow @rasbt](https://x.com/rasbt). Raschka explains language-model architecture, training, fine-tuning, and implementation. Follow him when a model release raises a deeper question such as what changed in attention, post-training, data, or inference. **Why follow:** his diagrams and code-oriented explanations help practitioners move past benchmark headlines. ### 5. Chip Huyen — @chipro [Follow @chipro](https://x.com/chipro). Huyen writes about designing and operating machine-learning and AI systems. Her feed is valuable for evaluation, data, product architecture, latency, cost, and the difference between a demo and a dependable system. **Why follow:** she consistently frames AI as a systems problem with business and operational constraints. ## Best AI Research Accounts ### Demis Hassabis — @demishassabis [Follow @demishassabis](https://x.com/demishassabis) for Google DeepMind research, models, and AI-for-science announcements. Treat launch posts as primary company context and pair them with independent testing. ### Fei-Fei Li — @drfeifei [Follow @drfeifei](https://x.com/drfeifei) for computer vision, spatial intelligence, human-centered AI, research institutions, and policy. ### François Chollet — @fchollet [Follow @fchollet](https://x.com/fchollet) for critical discussion of reasoning, generalization, intelligence, and benchmark design. His feed is especially helpful when consensus becomes too comfortable. ### David Ha — @hardmaru [Follow @hardmaru](https://x.com/hardmaru) for generative models, world models, research papers, and visually intuitive experiments. ### Jim Fan — @DrJimFan [Follow @DrJimFan](https://x.com/DrJimFan) for embodied AI, robotics, simulation, agents, and clear research demonstrations. ### Nathan Lambert — @natolambert [Follow @natolambert](https://x.com/natolambert) for open-weight models, reinforcement learning, post-training, and research analysis connected to his Interconnects writing. ## Best AI Builder and Engineering Accounts ### Andrew Ng — @AndrewYNg [Follow @AndrewYNg](https://x.com/AndrewYNg) for applied AI, education, data-centric development, and a measured view of how teams can adopt new techniques. ### Shawn Wang — @swyx [Follow @swyx](https://x.com/swyx) for the AI engineering ecosystem: agents, evaluations, developer tools, events, and emerging implementation patterns. Cross-check rapidly moving claims with the linked product documentation. ### Hamel Husain — @HamelHusain [Follow @HamelHusain](https://x.com/HamelHusain) for AI product evaluation, LLM engineering, fine-tuning, and practical lessons from helping teams ship systems. ### Shreya Shankar — @sh_reya [Follow @sh_reya](https://x.com/sh_reya) for data systems, evaluations, and rigorous thinking about building reliable AI applications. ### Harrison Chase — @hwchase17 [Follow @hwchase17](https://x.com/hwchase17) for primary updates and opinions from the LangChain and LangGraph ecosystem. Pair the feed with official changelogs before making implementation decisions. ## Official AI Lab Accounts to Follow Official accounts are the cleanest source for launches, documentation links, research publications, and availability notices: - [@OpenAI](https://x.com/OpenAI) - [@AnthropicAI](https://x.com/AnthropicAI) - [@GoogleDeepMind](https://x.com/GoogleDeepMind) - [@MetaAI](https://x.com/MetaAI) - [@MistralAI](https://x.com/MistralAI) - [@huggingface](https://x.com/huggingface) - [@AIatMeta](https://x.com/AIatMeta) - [@NVIDIAAI](https://x.com/NVIDIAAI) - [@Cohere](https://x.com/cohere) - [@perplexity_ai](https://x.com/perplexity_ai) Do not use an official lab account as your only source of analysis. It will accurately explain what the organization released, but it is still presenting that release from the organization's perspective. ## Best Accounts for AI Policy and Social Impact - [Stanford HAI — @StanfordHAI](https://x.com/StanfordHAI): research and policy from the Stanford Institute for Human-Centered AI - [NIST — @NIST](https://x.com/NIST): standards, risk management, measurement, and official U.S. technical guidance - [OECD.AI — @OECD_AI](https://x.com/OECD_AI): cross-country AI policy and data - [Ada Lovelace Institute — @AdaLovelaceInst](https://x.com/AdaLovelaceInst): governance and societal impacts - [AI Now Institute — @AINowInstitute](https://x.com/AINowInstitute): accountability, power, labor, and public-interest analysis The policy list should include sources with different institutional perspectives. Do not mistake repeated agreement inside one online network for broad consensus. ## The Accounts to Avoid An account does not deserve a place in your feed merely because it posts often. Be cautious when you see: - daily tool threads built primarily from affiliate links; - model comparisons with no prompts, settings, dates, or output evidence; - screenshots that omit the source account or original link; - anonymous leaks presented as confirmed releases; - benchmarks with no dataset, method, or reproducible artifact; - permanent urgency, where every update supposedly changes everything. Before reposting an AI claim, ask for the primary artifact: an official release, documentation page, research paper, repository, model card, evaluation method, or complete demonstration. If none exists, label the claim as unverified. ## Build a High-Signal AI X Feed Create four private Lists rather than relying on one home feed: 1. **Primary sources:** labs, research groups, standards bodies, and official product accounts 2. **Research:** people who publish or explain papers and evaluations 3. **Engineering:** builders who share code, system design, and real operating tradeoffs 4. **Work and policy:** educators, organizational researchers, and policy institutions Start with five to eight accounts per list. Review the lists separately when you have a reason: research on Monday, product releases during a launch, engineering when evaluating a tool. ### A Simple Quarterly Audit Keep an account if at least two of these are true: - It regularly links primary sources. - It adds expertise you cannot get from a lab announcement. - It shows methods, prompts, code, or limitations. - It separates fact, interpretation, and prediction. - It has corrected or updated an earlier claim. Mute or remove it if the feed is mostly outrage, affiliate promotion, screenshots without context, or repeated summaries of other people's work. ## Related Guides - [12 Best AI Subreddits in 2026, Ranked by Signal](/blog/best-ai-subreddits-for-discussion) - [10 Best AI Discord Servers for Networking in 2026](/blog/best-ai-discord-servers-for-networking) - [The Best AI Blogs and Websites for News](/blog/best-ai-blogs-and-websites-for-news) - [Best AI Newsletters to Subscribe To](/blog/best-ai-newsletters-to-subscribe-to) **Who is the best AI account to follow on X?** For a single general-purpose follow, start with Andrej Karpathy for technical intuition or Simon Willison for rapid, documented testing of new AI products. Add Ethan Mollick if your focus is workplace and education rather than engineering. **Is X still useful for AI news in 2026?** Yes, especially for primary announcements and discussion at release time. It is less reliable as an archive or verification layer. Use X for discovery, then follow links to official documentation, papers, repositories, and complete evaluations. **How many AI X accounts should I follow?** Start with about twenty split across primary sources, research, engineering, and work or policy. The exact number matters less than whether you can review the feed without missing evidence-bearing posts. **Should I follow AI labs or individual researchers?** Follow both. Lab accounts provide primary announcements and official links; researchers and independent builders add context, criticism, and hands-on testing. Neither group is sufficient alone. **How can I tell whether an AI influencer is reliable?** Look for primary links, disclosed methods, complete examples, clear uncertainty, and corrections. Be skeptical of accounts that publish constant superlatives, unexplained benchmarks, or affiliate-heavy tool lists. **Are follower counts a good way to rank AI accounts?** No. Follower counts measure reach, not accuracy or usefulness. A specialist who posts one careful evaluation per month may be far more valuable to your work than a large account that republishes every launch. ## Bottom Line Build a small portfolio of perspectives. Use official lab accounts to learn what shipped, Karpathy and Raschka for technical understanding, Willison and Huyen for evidence-driven implementation, Mollick and Ng for applied use, and policy institutions for the wider consequences. Then prune aggressively whenever an account stops earning attention.