Over the last three years, the field of artificial intelligence has changed faster than any other technology in history. GPT-4, Gemini, Claude, Llama, Sora, DeepSeek, Grok… launches keep coming at a pace that makes it almost impossible to keep up without a clear strategy.
In this article I show you exactly how to stay current: which newsletters to read, who to follow on X, which podcasts to listen to, and how to structure your week to absorb what matters without wasting hours on noise.
1. The Problem: Too Much AI, Too Little Time
The AI ecosystem generates hundreds of articles, papers, and announcements every week. Just arXiv publishes between 200 and 300 AI-related papers per day. Add to that the press releases from OpenAI, Google, Anthropic, Meta, Mistral, xAI… and the social media hype that accompanies every launch.
The Most Common Mistake
Trying to read everything. Most developers I know swing from ignoring AI entirely to trying to consume all available content, which leads to paralysis and burnout. The key is not to read more, but to curate better.
The goal of this guide is to give you a system of 30 to 45 minutes a day that keeps you informed about what truly matters: the models that will change your workflow, the tools that will make your work more efficient, and the trends that will define the market over the coming months.
2. The Best AI Newsletters
Newsletters are the most efficient way to consume AI news. A good editor has already done the curation for you: filtered out the noise, identified what matters, and summarized it into a 5-minute read.
The Batch
deeplearning.ai · Weekly
Andrew Ng's newsletter. Weekly technical and business synthesis. If you're only going to read one, make it this one. Excellent signal-to-noise ratio.
TLDR AI
TLDR Media · Daily
3 bullets per item, every day. Perfect for your morning coffee. Covers research, applications, and tools. High cadence without being overwhelming.
The Rundown AI
Daily · 700k+ subscribers
The most popular in the ecosystem. Very accessible, good for non-technical readers too. Covers both hype and practical tools.
Latent Space
swyx & Alessio · Weekly
For AI engineers. Deep, technical, with interviews with the builders of the ecosystem. If you build with LLMs, this is mandatory.
Import AI
Jack Clark · Weekly
From one of the co-founders of Anthropic. Focused on AI research and geopolitics. Long-term perspective, highly recommended.
Ben's Bites
Ben Tossell · Daily
Excellent for discovering new tools and practical use cases. Very focused on product and application, less on pure research.
My Recommended Personal Stack
I read TLDR AI every morning (5 min), The Batch on Wednesdays (15 min), and Latent Space on weekends when I want to go deeper. The rest when the topic is especially relevant.
3. X (Twitter) Accounts Worth Following
X is still the network where top AI researchers and builders post in real time. Building a good list is like having a curated feed of first-rate news.
🔬 Researchers & Scientists
@karpathy
Andrej Karpathy — ex-OpenAI, ex-Tesla. The best technical communicator in the ecosystem.
@ylecun
Yann LeCun — Chief AI Scientist at Meta. Always provides deep technical perspective and controversy.
@fchollet
François Chollet — creator of Keras. Very critical and rigorous about hype.
@hardmaru
David Ha — ex-Google Brain. Focused on creativity and generative AI.
@drjimfan
Jim Fan — NVIDIA Research. Excellent threads on agents and embodied AI.
@GaryMarcus
Gary Marcus — The smartest critic of the hype. Useful for calibrating expectations.
⚒️ Builders & Developers
@simonw
Simon Willison — creator of Datasette. The best practical guides on LLMs.
@swyx
Shawn Wang — co-host of Latent Space. Very active in the tools ecosystem.
@emollick
Ethan Mollick — Wharton professor. The best at showing practical AI applications.
🏢 Official Accounts
@AnthropicAI
Claude / MCP updates
@OpenAI
GPT / o-series
@GoogleDeepMind
Gemini / research
@AIatMeta
Llama / research
@MistralAI
Open source models
@huggingface
Open source ecosystem
Tip: Create a Private List
Instead of following everyone and polluting your main feed, create a private list on X with these accounts and check it 1-2 times a day. Separate intentional consumption from casual scrolling.
4. Podcasts and YouTube Channels
For audio and video formats, the key is to integrate them into activities you already do: exercise, commuting, cooking. No need to carve out extra time.
🎙️ Podcasts
Latent Space Podcast
The best for engineers. Long-form interviews with the builders of the ecosystem. swyx & Alessio.
Lex Fridman Podcast
2-4 hour interviews with key figures. When it's about AI, it's excellent. Be selective.
No Priors
Interviews with founders and CEOs of the most important companies in the AI ecosystem.
Hard Fork (NYT)
Analysis of technology and AI from a broader perspective. Good for understanding the social impact.
TWIML AI Podcast
This Week in Machine Learning. Technical, with researchers. One of the most long-standing and reliable.
📺 YouTube
Andrej Karpathy
His tutorials are the gold standard for understanding LLMs from first principles. Mandatory.
Two Minute Papers
Research paper summaries in 3-5 minutes. Ideal for staying current without reading full papers.
AI Explained
Deep, balanced analysis of the most important launches. No unnecessary hype.
Matt Wolfe
For following tools and practical applications. Very accessible and high publication frequency.
5. Tools for Aggregating and Filtering
Beyond following individual sources, there are platforms that automatically aggregate and organize AI content.
Hacker News
The premier tech community. Filter by "AI" or "LLM" and you'll find the smartest discussions around every launch. Use hn.algolia.com to search for specific topics.
Papers With Code
If you want to follow the state of the art in research, this is the best source. Shows the latest papers with available implementations. The "Trending" section is essential.
Perplexity AI
For querying recent news with context. "What happened with [model X] this week?" It's more efficient than Google searches for catching up quickly.
Hugging Face Daily Papers
Daily curation of the most relevant arXiv papers, voted on by the community. Available at huggingface.co/papers. Ideal for following research without getting lost in arXiv.
RSS with Feedly / Inoreader
If you prefer to centralize sources, building an RSS reader with the blogs of OpenAI, Anthropic, Google DeepMind, Meta AI, and Mistral gives you a unified view of all official announcements without depending on algorithms.
6. A Practical Routine That Works
The key is not the number of sources but consistency. This is the structure that works best for me:
☀️ Daily (15–20 minutes)
- 1. Read TLDR AI while having your morning coffee (5 min)
- 2. Check the X list with key researchers (5–10 min)
- 3. If something is worth going deeper on, save it to Pocket / Notion for later
📅 Weekly (1–2 hours)
- Wednesday: Read the full Andrew Ng's The Batch (15 min)
- Thursday/Friday: One episode of Latent Space Podcast during exercise
- Weekend: Read a long article or paper saved during the week
📆 Monthly
- Week 1: Check Papers With Code "Trending" to see what research is gaining traction
- Week 2: Clean up subscriptions — unsubscribe from what you don't read and subscribe to new relevant sources
- Week 4: Write a personal summary of the most important changes of the month (helps with retention)
7. How to Separate Signal from Noise
80% of AI content published is hype, repetition, or superficial analysis. Here are the signals that help identify what is worth reading:
✅ Worth reading
- • Direct publications from the labs (OpenAI, Anthropic, Google)
- • Papers with available code on Papers With Code
- • Comparative benchmarks with clear methodology
- • Analysis from @karpathy, @simonw, or @emollick on a topic
- • Open source projects that surpass 1k GitHub stars in 48 hours
- • Discussions with more than 100 comments on Hacker News
❌ Usually noise
- • "This will replace [job X] forever"
- • Benchmark comparisons without real-world usage context
- • Twitter threads from people without a technical track record
- • Articles from generalist media about AI without a technical author
- • Spectacular demos without public access or a paper
- • Predictions more than 6 months out in this field
Golden rule
If an announcement is truly important, you'll see it confirmed by multiple reliable sources within 24–48 hours. You don't need to be the first to find out. Better to read less and read well than to skim everything superficially.
Conclusion
Staying current with AI doesn't require being glued to screens all day. It requires smart curation: choosing sources wisely, building a consistent routine, and learning to distinguish what's important from what's merely urgent.
The stack I've proposed — TLDR AI + The Batch + Latent Space + curated X list — gives you comprehensive coverage in under 45 minutes a day. The rest is going deeper based on your specific interests.
Recommended Stack: Summary
Newsletters
- • TLDR AI (daily)
- • The Batch (weekly)
- • Latent Space (weekly)
Networks & Aggregators
- • Curated X list
- • Hacker News / AI filter
- • HuggingFace Papers
Audio / Video
- • Latent Space Podcast
- • Two Minute Papers
- • AI Explained (YT)
