AI collection

The Best AI Newsletters in 2026

Twelve AI newsletters that are still publishing, sorted by what kind of reader each one actually serves.

Most AI newsletters fall into two groups. Daily digests like The Rundown AI and TLDR AI tell you what shipped yesterday in a five-minute read. Weekly or irregular essays like Import AI, Interconnects and One Useful Thing pick one or two ideas and go deep. If you only have time for one, decide whether you need to keep up with announcements or understand what they mean.

A simple setup works for most people: one daily digest for the headlines, and one or two writers whose judgment you trust for the slower thinking. Builders should add something practical like Simon Willison or Latent Space. If the daily volume starts to feel like noise, drop the digest before you drop the essayists.

At a glance

#NewsletterBest forCadencePrice
1The Batchengineers and managers who want a calm weekly summaryWeeklyFree
2Import AIreaders who want research papers and policy in one placeRoughly weeklyFree
3The Rundown AIbusy professionals who want AI news in five minutesDailyFree
4TLDR AIdevelopers who want technical AI links, not hypeDaily (weekdays)Free
5Ben's Bitesnon-engineers building with AI agents and toolsSeveral issues a weekFree
6Last Week in AIreaders who prefer one weekly catch-up over daily emailsWeeklyFree, paid tier available
7Interconnectstechnical readers following open models and trainingSeveral posts a weekFree, paid tier available
8Latent SpaceAI engineers building products on top of modelsSeveral posts a weekFree, paid tier available
9One Useful Thingmanagers, educators and knowledge workersEvery few weeksFree, paid tier available
10Simon Willison's Newsletterdevelopers who use LLMs every dayRoughly weeklyFree
11AI as Normal Technologyskeptics and policy readers who want evidence over predictionsEvery few weeksFree
12The Prompt Warriorsmall business owners automating their own workRoughly weeklyFree (issues are public in the archive)

The list

No.01

The Batch

By DeepLearning.AI (Andrew Ng)
WeeklyFreeOwn siteBest for engineers and managers who want a calm weekly summary

Each issue opens with a letter from Andrew Ng, so you get a practitioner's take along with the news.

A weekly roundup of the main AI news stories, each explained in a few paragraphs with why it matters. Issues start with a short letter from Andrew Ng on a topic he is thinking about.

No.02

Import AI

By Jack Clark
Roughly weeklyFreeSubstackBest for readers who want research papers and policy in one place

Jack Clark summarizes research papers most outlets skip and adds his own read on what they mean for policy and safety.

Numbered issues that walk through several recent papers or technical results, each with a short explanation and commentary. Most issues end with a piece of short fiction about AI.

No.03

The Rundown AI

By Rowan Cheung and The Rundown team
DailyFreebeehiivBest for busy professionals who want AI news in five minutes

A fast daily scan of the main AI stories, written for people who use AI at work rather than build it.

Each daily issue covers a handful of AI news stories with a short summary and why it matters, plus tool picks and practical tips.

Checked active Sep 29, 2026. Recent issue: OpenAI's agents went rogue on Washington (Sep 28, 2026)
2,000,000+ readers (per therundown.ai)Visit
No.04

TLDR AI

By TLDR
Daily (weekdays)FreeOwn siteBest for developers who want technical AI links, not hype

Short, link-first summaries that lean toward research, models and engineering rather than business gossip.

A daily email of AI links, each with a two or three sentence summary, grouped into headlines, research, engineering and quick links.

1,100,000 readers (per tldr.tech)Visit
No.05

Ben's Bites

By Ben Tossell
Several issues a weekFreeSubstackBest for non-engineers building with AI agents and tools

Ben writes from his own experiments building with agents, so the tool picks come with first-hand notes.

Short posts on what Ben is learning while building with AI agents, plus tools and launches worth trying.

Checked active Sep 29, 2026. Recent issue: 50 years of tech devices (Sep 25, 2026)
Over 171,000 subscribers (per bensbites.com)Visit
No.06

Last Week in AI

By Last Week in AI team
WeeklyFree, paid tier availableSubstackBest for readers who prefer one weekly catch-up over daily emails

One weekly issue sorts the week's AI news by category, so you can catch up without reading every day.

A weekly text summary of the week's AI news, grouped into sections such as tools, business, research and policy, with a companion podcast.

Over 184,000 subscribers (per lastweekin.ai)Visit
No.07

Interconnects

By Nathan Lambert
Several posts a weekFree, paid tier availableSubstackBest for technical readers following open models and training

Nathan Lambert trains models himself, so his posts on open models and post-training are grounded in hands-on work.

Essays and interviews on how frontier and open models are built and trained, the state of open-weight models, and the AI research landscape.

No.08

Latent Space

By swyx and Alessio Fanelli
Several posts a weekFree, paid tier availableSubstackBest for AI engineers building products on top of models

It covers how AI products are actually built, from the people building them, which few news digests do.

Podcast interviews with founders and researchers published with show notes, essays on AI engineering, and the AINews recap of what people are discussing.

Over 201,000 subscribers (per latent.space)Visit
No.09

One Useful Thing

By Ethan Mollick
Every few weeksFree, paid tier availableSubstackBest for managers, educators and knowledge workers

Ethan Mollick tests AI tools on real work and explains the results for people who are not engineers.

Longer essays on how AI is changing work and education, often based on his own tests of the latest models, including a regularly updated guide to which AI to use.

Checked active Sep 29, 2026. Recent issue: The Overhang (Sep 18, 2026)
Over 479,000 subscribers (per oneusefulthing.org)Visit
No.10

Simon Willison's Newsletter

By Simon Willison
Roughly weeklyFreeSubstackBest for developers who use LLMs every day

Simon tries new models and tools the day they come out and writes up exactly what worked, with code and prompts.

A roundup of the posts from Simon's blog: hands-on notes on new models, LLM tooling, prompt injection and security, and his own open source projects.

Checked active Sep 29, 2026. Recent issue: 2026 in LLMs (so far) (Sep 28, 2026)
Over 71,000 subscribers (per simonw.substack.com)Visit
No.11

AI as Normal Technology

By Arvind Narayanan and Sayash Kapoor
Every few weeksFreeSubstackBest for skeptics and policy readers who want evidence over predictions

Two Princeton researchers (formerly publishing as AI Snake Oil) check big AI claims against the evidence.

Essays that examine claims about AI capabilities, risk and impact, arguing that AI is a transformative but normal technology. The newsletter was previously called AI Snake Oil.

No.12

The Prompt Warrior

By Moritz Kremb
Roughly weeklyFree (issues are public in the archive)beehiivBest for small business owners automating their own work

Short, practical issues on the exact AI workflows Moritz uses to run his business.

Brief weekly emails with a few AI workflows, tools or automations he is using, plus occasional beginner guides such as using Claude Code without coding experience.

Questions

AI newsletters: FAQ

What is the best AI newsletter for beginners?

One Useful Thing and The Rundown AI are the easiest starting points. One Useful Thing explains what AI means for everyday work, and The Rundown AI gives a short daily summary of the news without assuming technical knowledge.

Which AI newsletter is best for developers?

Simon Willison's newsletter, Latent Space and TLDR AI are the most useful for people who build with models. Simon shares hands-on tests, Latent Space covers how AI products are built, and TLDR AI collects technical links each weekday.

Are AI newsletters free?

All twelve on this list can be read for free. Several, including Interconnects, Latent Space and One Useful Thing, offer an optional paid tier, but most of their issues are free to read.

How many AI newsletters should I subscribe to?

Two or three is usually enough: one daily or weekly digest for news, and one or two essay writers for analysis. More than that tends to repeat the same stories.