- Home
- Best AI newsletters
The 21 Best AI Newsletters
There are now more AI newsletters than there were AI companies five years ago. Most "best AI newsletters" lists are written to sell you the author's own letter, so the ranking is decided before the research starts. This page is different in two honest ways. First, we publish The Agent Signal, a daily AI newsletter family with 23 lanes, so we are a competitor to everyone on this list, and we say so. Second, we only write down what each newsletter states about itself on its own site, with the link, so you can check us in one click.
The list covers the daily AI news letters most people start with (The Rundown, TLDR AI, Superhuman, Ben's Bites, The Neuron), the weekly deep reads that researchers and engineers actually forward (Import AI, The Batch, Ahead of AI, Interconnects, Latent Space, Last Week in AI, AlphaSignal), the regional specialist (ChinaTalk), and seven of our own lanes. Each card follows the same template: what it is, cadence, price, who it is best for, and our honest take. Where a newsletter beats ours for a particular reader, we say that too.
How we picked and ranked
We did not rank by subscriber count. Self-reported counts are unauditable and reward whoever bought the most ads. We ranked by a simpler question: if a reader tells us what they want from an AI newsletter, which one would we honestly send them to first? That produces a rough order — general daily news at the top, specialist weeklies in the middle, our own lanes at the end so nobody accuses us of burying the competition under our brand.
Facts were gathered on 2026-09-12 from each newsletter's homepage. We fetched ten sites; The Neuron returned an HTTP 403 to our fetch, and we did not fetch Ahead of AI, ChinaTalk, or AI Breakfast this pass, so those four cards carry an UNVERIFIED label and no numbers. We do not use Reddit threads, scraped search-engine snippets, or reviews we cannot attribute to a named person.
The list at a glance
- The Rundown AI
- TLDR AI
- Superhuman AI
- Ben's Bites
- The Neuron
- Last Week in AI
- Import AI
- The Batch
- AlphaSignal
- Ahead of AI
- Latent Space
- Interconnects
- ChinaTalk
- AI Breakfast
- Agent Signal News (ours)
- The AI Agent Stack (ours)
- Claude Agent Signal (ours)
- OpenAI Agent Signal (ours)
- Gemini Agent Signal (ours)
- Open Weights (ours)
- The China Agent Signal (ours)
Daily AI news newsletters
These are the letters that try to be your five-minute morning read. They overlap heavily — on a big launch day, all five lead with the same story — so the difference is voice, tool picks, and how much they respect your time.
1. The Rundown AI
- What it is
- A daily AI news and tools letter that promises to teach you AI "in 5 minutes a day", with story cards, guides and a trending-tools section. Built on beehiiv.
- Cadence
- Daily
- Price
- Free (paid tiers exist for courses and community; not stated on the homepage we fetched)
- Author
- Rowan Cheung and team
- Best for
- Anyone who wants the headline, one sentence of why it matters, and a tool to try — with zero technical depth required.
- Source
- therundown.ai (homepage fetched 2026-09-12; meta description self-reports "2,000,000+ readers")
Our take. The Rundown is the default recommendation for a reason: it is consistent, fast, and the formatting is the cleanest in the category. What it is not is analytical. Stories are summarised, not argued, and the tools section leans toward whatever is trending rather than what is durable. If you already know what an LLM is and want to know what changed for your work, you will outgrow it within a few months — but it is the right first subscription for most people.
2. TLDR AI
- What it is
- The AI edition of the TLDR family — a free daily newsletter of AI news, research breakthroughs and developer tools in roughly five minutes of reading. Link-heavy, minimal commentary.
- Cadence
- Daily on weekdays
- Price
- Free
- Author
- TLDR editorial team (individual editors not named on the signup page)
- Best for
- Engineers and technical readers who want links with one-line summaries and no opinions.
- Source
- tldr.tech/ai (fetched 2026-09-12; page self-reports "1,100,000 subscribers")
Our take. TLDR AI is the most efficient letter on this list: each item is a headline, a reading time and a two-sentence summary. It respects your attention more than anyone. The trade-off is that it never tells you what to think, and its research section skews toward arXiv abstracts you may not have context for. If you want a curated link dump from people who read a lot, this is it. If you want a narrative, look elsewhere.
3. Superhuman AI
- What it is
- A daily newsletter whose pitch is "Get the latest AI news and tools, and learn how to use them at work." Strong on step-by-step prompts and workplace use cases.
- Cadence
- Daily
- Price
- Free
- Author
- Zain Kahn, Theodore and the Superhuman AI team (as stated on the site)
- Best for
- Office workers and managers who want AI as a productivity tool rather than a technology story.
- Source
- superhuman.ai (fetched 2026-09-12; the page self-reports both "2 million+ pros" and "1 million+ readers" in different sections — we note the inconsistency rather than pick one)
Our take. Superhuman's edge is the "how to use it at work" framing: most issues include an actual prompt or workflow. Its weakness is that the news coverage is thin and sometimes a day behind; it is a training letter dressed as a news letter. That is fine if you want training. For people who need to know what happened today, it is a second subscription, not a first.
4. Ben's Bites
- What it is
- Ben Tossell's letter, which started as a daily AI news round-up in 2022 and has shifted toward agent-building insights and tool recommendations, per its current homepage.
- Cadence
- Not stated on the homepage we fetched
- Price
- Free to subscribe (no pricing shown on the homepage)
- Author
- Ben Tossell
- Best for
- Makers and indie builders who want a founder's voice and hands-on tool tests.
- Source
- bensbites.com (fetched 2026-09-12; page self-reports "over 170,000 subscribers")
Our take. Ben's Bites reads like a friend telling you what happened yesterday, and that voice is the whole product. The recent pivot toward building with agents makes it more useful to us than the old news digest was. It is not comprehensive and does not pretend to be. Subscribe for the taste, not the coverage.
5. The Neuron
- What it is
- A daily AI newsletter known for a light, humorous tone and tool picks. UNVERIFIED: the site returned HTTP 403 to our fetch on 2026-09-12, so cadence, price and audience are not confirmed here.
- Cadence
- UNVERIFIED (reported daily)
- Price
- UNVERIFIED (reported free)
- Author
- UNVERIFIED
- Best for
- Readers who find the other dailies too dry.
- Source
- theneurondaily.com (fetch blocked; details unverified)
Our take. We have read The Neuron and like it, but we will not put numbers on a card we could not verify today. If you want the same news as The Rundown with more jokes, this is the one. We will refresh this card on the next review pass.
Weekly deep reads
If dailies are the morning paper, these are the weekend magazine. They publish less, argue more, and are written by people who work in or next to the labs. For anyone doing serious work in AI, at least two of these belong in your inbox.
6. Last Week in AI
- What it is
- "Weekly text and audio summaries of the most interesting AI news, as well as editorials commenting on recent events" — a newsletter paired with a long-running podcast.
- Cadence
- Weekly
- Price
- Free
- Author
- Not named on the homepage we fetched
- Best for
- People who prefer to listen; commuters; anyone who wants one weekly catch-up instead of five dailies.
- Source
- lastweekin.ai (fetched 2026-09-12; page self-reports "over 180,000 subscribers")
Our take. The text edition is a well-organised link list grouped by theme; the podcast is where the opinion lives. It is the best single weekly recap for a generalist. It is not fast — by design — so if a story matters to your job on the day it breaks, you need a daily as well.
7. Import AI
- What it is
- Jack Clark's long-running "weekly newsletter about artificial intelligence based on detailed analysis of cutting-edge research." Papers, policy, and a short fiction piece at the end.
- Cadence
- Weekly
- Price
- Free
- Author
- Jack Clark (co-founder of Anthropic; the newsletter is his personal publication)
- Best for
- Policy people, researchers, and anyone who wants to understand what the frontier labs are worried about.
- Source
- importai.substack.com (fetched 2026-09-12; page self-reports "over 140,000 subscribers")
Our take. Import AI is the most-forwarded letter among people who work on AI policy, and it earns that. Each paper summary ends with "Why this matters", which is the whole discipline of newsletter writing in three words. It is dense and assumes you can read a chart. It is also, obviously, written by someone inside one lab, and readers should weigh that. Our AI Safety Signal and Frontier AI Research lanes cover the same territory daily and with more sources, but Import AI has the deeper voice.
8. The Batch
- What it is
- DeepLearning.AI's weekly, with an opening letter from Andrew Ng followed by several news and research items with clear "why it matters" sections.
- Cadence
- Weekly
- Price
- Free
- Author
- DeepLearning.AI editorial team; Andrew Ng writes the opening letter
- Best for
- Learners and practitioners who want the educator's view — what a story means for someone building skills.
- Source
- deeplearning.ai/the-batch (fetched 2026-09-12; no reader count stated)
Our take. The Batch is the most carefully edited letter on this list. Every item has context, and the tone is calm in a category that runs on hype. Its limits are pace (weekly) and a natural bias toward DeepLearning.AI's own courses. If you are learning AI rather than shipping it, start here alongside our beginner lane, The AI Shortcut.
9. AlphaSignal
- What it is
- A developer-facing letter and site whose pitch is to "track every model, repo, paper, and announcement across the AI industry in real time." Heavy on GitHub repos and trending papers.
- Cadence
- Not stated on the homepage we fetched
- Price
- Free, with "Pro" content noted on the site
- Author
- Founded by Lior Alexander; editorial team listed as Ben Dickson (editorial lead), Akruti Acharya, Adham Khaled and Rehman Rind
- Best for
- ML engineers who want repos and papers, not business news.
- Source
- alphasignal.ai (fetched 2026-09-12; no reader count stated on the page)
Our take. AlphaSignal is what TLDR AI would be if it only cared about code. The repo picks are genuinely useful and often a day ahead of the mainstream dailies. The writing is functional rather than good. It competes most directly with our Open-Source AI Agents and Open Weights lanes; AlphaSignal is broader across ML, ours are narrower on agents and runnable models.
10. Ahead of AI
- What it is
- Sebastian Raschka's Substack on LLM research and engineering — long, tutorial-style essays on architectures, training and evaluation. UNVERIFIED: not fetched this pass.
- Cadence
- UNVERIFIED (irregular, roughly monthly essays)
- Price
- UNVERIFIED (free posts with an optional paid tier)
- Author
- Sebastian Raschka
- Best for
- Engineers who want to understand how a model works, not what a company announced.
- Source
- magazine.sebastianraschka.com (not fetched; details unverified)
Our take. This is a textbook delivered in chapters, not a newsletter in the news sense. There is nothing on this list that teaches transformer internals as well. It is also slow and silent for weeks at a time, so pair it with something daily.
11. Latent Space
- What it is
- "The AI Engineer newsletter + top technical AI podcast. How leading labs build Agents, Models, Infra, & AI for Science." Newsletter, podcast and the AI Engineer conference series share the brand.
- Cadence
- Not stated on the homepage we fetched
- Price
- Free (paid tier exists on Substack; not shown on the page we fetched)
- Author
- swyx and Alessio (first names as used on the site)
- Best for
- Working AI engineers — people building with models rather than training them.
- Source
- latent.space (fetched 2026-09-12; page self-reports "over 200,000 subscribers")
Our take. Latent Space coined the "AI engineer" category and still defines it. The interviews are the best technical conversations in the space; the written recaps are thorough to the point of being long. It is closest in spirit to our AI Agent Stack lane — both are for builders deciding what to ship — but Latent Space is a community with a newsletter attached, and ours is a daily brief.
12. Interconnects
- What it is
- Nathan Lambert's letter: "The cutting edge of AI, from inside the frontier AI labs, minus the hype." Post-training, RLHF, open models and lab strategy.
- Cadence
- Not stated on the homepage we fetched (roughly weekly in practice — UNVERIFIED)
- Price
- Free, with a paid tier on Substack
- Author
- Nathan Lambert
- Best for
- Researchers and technical leads who want the open-model and post-training picture explained by someone who does it.
- Source
- interconnects.ai (fetched 2026-09-12; page self-reports "over 82,000 subscribers")
Our take. Interconnects is the most opinionated technical letter here and the most reliable on open-weight models — which is exactly the beat our Open Weights lane covers daily. Read Interconnects for the argument, ours for the changelog. Its sizes and rankings of models are researcher judgements, not benchmarks, and it is clear about that.
Specialist and regional
13. ChinaTalk
- What it is
- Jordan Schneider's newsletter and podcast on China, technology and policy, with substantial coverage of Chinese AI labs and export controls. UNVERIFIED: not fetched this pass.
- Cadence
- UNVERIFIED (multiple posts per week)
- Price
- UNVERIFIED (free posts with a paid tier)
- Author
- Jordan Schneider and contributors
- Best for
- Anyone tracking Qwen, DeepSeek and the policy fight around chips.
- Source
- chinatalk.media (not fetched; details unverified)
Our take. ChinaTalk is the deepest English-language source on the Chinese AI ecosystem, with translated interviews and policy analysis that nobody else does. It is a policy publication first and an AI newsletter second. Our China Agent Signal is the daily lab-by-lab tracker; ChinaTalk is the context you need to understand why those releases matter.
14. AI Breakfast
- What it is
- A concise AI news digest on beehiiv, positioned as a quick morning read. UNVERIFIED: not fetched this pass.
- Cadence
- UNVERIFIED
- Price
- UNVERIFIED
- Author
- UNVERIFIED
- Best for
- Readers who want a shorter alternative to the big dailies.
- Source
- aibreakfast.beehiiv.com (not fetched; details unverified)
Our take. We include AI Breakfast because readers ask about it, not because we can vouch for it today. Everything on this card will be verified or removed on the next review.
Our lanes — The Agent Signal
Disclosure: we publish these. We have not put a subscriber number on any of them because we do not think unaudited counts belong in a comparison, ours included. Every lane is a daily AI newsletter built the same way: every story checked against its source before anything ships, with a "why it matters" line on every story. Pricing is $9.99 a month for a single lane or $19.99 a month for the variety pass (all lanes). The full list of 23 lanes is below; these seven are the ones most people compare against the letters above.
Vendor names (Anthropic, Claude, OpenAI, Google, Gemini, xAI, Grok) are used nominatively to describe what a lane covers. The Agent Signal is not affiliated with or endorsed by any of these companies.
15. Agent Signal News
- What it is
- The flagship. "The news — smart, fast, zero fluff." A broadcast-structured daily led by what the whole industry is actually reporting, not what is loudest.
- Cadence
- Daily
- Price
- $9.99/month, or included in the $19.99 variety pass
- Author
- Harnoor Minhas, editor; how every issue is built and checked is described at How we make it
- Best for
- Anyone building or working with AI who wants the day's signal in five minutes with an "AI agents" lens.
- Subscribe
- Agent Signal News hub
Our honest take. Against The Rundown and TLDR AI, our advantage is depth of sourcing and the agents angle: we track hundreds of sources across the industry, and we tell you what to do with each one. Our disadvantage is history — we are the new letter on this page, our archive is short, and our voice is still settling. If you want the safe choice, pick The Rundown. If you want the agents-first daily, pick this and hold us to it.
16. The AI Agent Stack
- What it is
- "Your deep dive into agentic AI" — architectures, vendor strategy and the meta-story across the ecosystem, for the builder deciding what to ship and what to hedge.
- Cadence
- Daily
- Price
- $9.99/month, or included in the variety pass
- Author
- Harnoor Minhas
- Best for
- Engineers and technical founders building agent products.
- Subscribe
- AI Agent Stack hub
Our honest take. This is the lane closest to Latent Space and Interconnects, and it is daily where they are not. It is also less personal than either — it is a brief, not an essay — and a reader who wants a person's voice should subscribe to one of those as well. Read What is an AI agent? first if the vocabulary is new.
17. Claude Agent Signal
- What it is
- "Inside Anthropic & Claude — models, research, safety." A single-vendor daily: model releases, Claude Code, research papers, safety disclosures.
- Cadence
- Daily
- Price
- $9.99/month, or included in the variety pass
- Author
- Harnoor Minhas
- Best for
- Teams standardised on Claude who need every change on the day it lands.
- Subscribe
- Claude Agent Signal hub · Not affiliated with Anthropic
Our honest take. No general newsletter on this list tracks a single lab daily; that is the gap this fills. The honest limit is that on a quiet week a single-vendor lane is thin, and we say so in the issue rather than pad it.
18. OpenAI Agent Signal
- What it is
- "Everything OpenAI — models, Sora, ChatGPT." Releases, pricing changes, API deprecations, policy.
- Cadence
- Daily
- Price
- $9.99/month, or included in the variety pass
- Author
- Harnoor Minhas
- Best for
- Developers and operators whose stack depends on OpenAI.
- Subscribe
- OpenAI Agent Signal hub · Not affiliated with OpenAI
Our honest take. OpenAI generates more headlines than any other lab, which makes this lane the easiest to fill and the hardest to keep honest. Leading with what independent outlets actually confirmed is what stops it turning into a press-release feed.
19. Gemini Agent Signal
- What it is
- "Google DeepMind, Workspace & Gemini, daily." Models, Workspace integrations, Vertex and research.
- Cadence
- Daily
- Price
- $9.99/month, or included in the variety pass
- Author
- Harnoor Minhas
- Best for
- Google-first shops and anyone tracking DeepMind research.
- Subscribe
- Gemini Agent Signal hub · Not affiliated with Google
Our honest take. Google ships across so many surfaces that the general dailies only catch the headline model. This lane catches the Workspace and Vertex changes too. Its risk is the opposite of thinness: too much, so the ranking matters.
20. Open Weights
- What it is
- "Open-source & open-weight models you can run yourself, daily." New checkpoints, licences, quantisations, inference tooling.
- Cadence
- Daily
- Price
- $9.99/month, or included in the variety pass
- Author
- Harnoor Minhas
- Best for
- Anyone running models locally or on their own cloud.
- Subscribe
- Open Weights hub
Our honest take. Interconnects is the better essay on open models; this is the better daily changelog. We flag licence terms on every release because "open" on a model card does not always mean open for your use.
21. The China Agent Signal
- What it is
- "China's AI labs, daily — Qwen, DeepSeek, Baidu, Tencent, Moonshot, Zhipu." Lab-by-lab releases and benchmarks, in English.
- Cadence
- Daily
- Price
- $9.99/month, or included in the variety pass
- Author
- Harnoor Minhas
- Best for
- Readers who noticed that half the strongest open models now come from Chinese labs and want a daily tracker.
- Subscribe
- China Agent Signal hub
Our honest take. ChinaTalk has the policy depth and the interviews; we have the daily cadence and the model-release detail. If you only read one, read ChinaTalk. If you ship on Qwen or DeepSeek, read both.
All 23 Agent Signal lanes
Counting note: the ranking above has 21 entries — 14 newsletters from other publishers and 7 of our own lanes. The grid below is the full list of 23 Agent Signal lanes; the 16 lanes not ranked above are listed here for completeness, not as part of the "21 best".
Every lane is a daily AI newsletter with its own hub page and archive. The four training lanes are hands-on lessons rather than news.
- Agent Signal NewsThe news — smart, fast, zero fluff
- The AI Agent StackDeep dive into agentic AI
- Claude Agent SignalAnthropic & Claude · not affiliated
- Gemini Agent SignalGoogle DeepMind & Gemini · not affiliated
- OpenAI Agent SignalEverything OpenAI · not affiliated
- Grok Agent SignalxAI & Grok · not affiliated
- Open WeightsModels you can run yourself
- The China Agent SignalQwen, DeepSeek, Baidu, Tencent, Moonshot, Zhipu
- Open-Source AI AgentsFrameworks, MCP, orchestration
- Agentic AI EdgeAgents, frameworks, coding tools
- The AI OperatorFunding, strategy, policy
- Hyperscale Cloud AIAWS/Azure/GCP, inference, MLOps
- The AI Chip FoundryGPUs, NPUs, accelerators
- Embodied AI RobotsHumanoids, embodied agents
- AI Safety SignalAlignment, red-teaming, regulation
- Creative Agent SignalImage, video, audio, 3D
- AI at WorkLLMOps & enterprise productivity
- Frontier AI ResearchPapers, benchmarks, architectures
- The AI ShortcutAI for beginners
- OpenAI TrainingHands-on lessons
- NVIDIA TrainingHands-on lessons · not affiliated
- Cloud TrainingHands-on lessons
- AI/ML TrainingHands-on lessons
Which AI newsletter should you pick?
| If you are… | Start with | Then add |
|---|---|---|
| New to AI and want the news without jargon | The Rundown or The AI Shortcut | The Batch |
| An engineer who wants links, not opinions | TLDR AI | AlphaSignal |
| Building agent products | The AI Agent Stack | Latent Space |
| Standardised on one vendor | The matching lane: Claude, OpenAI, Gemini, Grok | The Rundown for the wider picture |
| Running open models | Open Weights | Interconnects |
| Working in policy or safety | Import AI | AI Safety Signal |
| Watching China's labs | ChinaTalk | The China Agent Signal |
| Prefer audio, once a week | Last Week in AI | Any daily for breaking days |
| Want AI as a work tool | Superhuman AI | AI at Work |
How to judge any AI newsletter (including ours)
Four tests, all checkable in one issue.
1. Does every story link to a primary source? A newsletter that links to another newsletter's summary is a copy of a copy. The Batch, Import AI and TLDR AI pass this cleanly. We require a source link on every card.
2. Is there a "why it matters" line, and is it specific? "This is huge for the industry" is not a why. "This drops the price of long-context calls below the point where RAG is worth building" is. Import AI set the standard here.
3. Are the reader numbers consistent? We found one homepage claiming two different audience sizes in different sections. It does not make the newsletter bad; it does tell you what the number is for. Treat every count on this page, ours included, as marketing.
4. What happens on a slow day? The best letters shorten. The worst ones pad. Read one issue from a quiet week before you judge a letter by its launch-day coverage.
Free vs paid AI newsletters: what you are actually paying for
Almost every daily on this list is free to the reader, which means the reader is not the customer. A free daily AI newsletter with a large list is an advertising product: the sponsor slot near the top, the "presented by" line, and the occasional issue that is a single sponsor's story are how the bills get paid. That is not a criticism — most good trade press has always worked this way — but it changes how to read the issue. Treat the top slot as an ad even when it is written in the editorial voice, and treat product roundups with a suspicious number of affiliate-style links the same way.
Paid newsletters flip the incentive. When you are the customer, the writer's job is to keep you, not to keep a sponsor, and the letter can be narrower, more opinionated and less frequent. The weeklies that charge (or run a paid tier) tend to be written by one identifiable person with a track record, and that person's reputation is the product. The trade-off is that you are buying one point of view, so a paid weekly pairs best with a free daily rather than replacing it.
Three questions settle whether a paid tier is worth it for you. First, is the paid content different in kind, or just more of the same? A paid tier that offers an archive, a community and "extra issues" is selling access; one that offers original analysis you cannot get elsewhere is selling judgement. Second, does the writer disclose conflicts — investments, advisory roles, employers? A paid letter with undisclosed positions is worse than a free one with ads. Third, can you cancel from the same page you subscribed on? If not, the publisher has told you something about how they see the relationship.
Our own lanes are paid; the price is listed in the FAQ below. We say this here because a comparison page that hides its own business model fails its own test.
How to set up your inbox so the newsletters actually get read
The most common outcome of subscribing to five AI newsletters is that none of them get read after week two. The problem is rarely the letters; it is that they land in the same inbox as everything else and get triaged like everything else. A few habits fix this for most people.
- Route them out of the main inbox. One filter — sender contains "newsletter", or a label per sender — moves every issue into a folder that you open on purpose. Unread counts in the main inbox are the enemy of reading.
- Pick a reading slot, not a reading goal. Ten minutes with coffee, or the first ten minutes after lunch, beats "I'll read them when I have time". A daily letter is designed to be finished in that slot; a weekly is designed for a longer one at the weekend.
- Use RSS if you already use a reader. Many newsletters publish a web version with a feed, and a reader keeps them out of email entirely. Every Agent Signal lane has its own RSS feed on its page, and there is a site-wide feed for all lanes.
- Apply the two-week rule. If you have not opened a letter in two weeks, unsubscribe. You can always come back; an unread pile is a small daily tax on attention, and the letters you keep will be read more carefully.
- Keep one, not five, for the same job. Two daily news letters covering the same stories is a duplicate, not a second opinion. Pick the one whose format you finish, and add letters only for a different job — a weekly for depth, a specialist lane for your vendor or sector.
Signs a newsletter is worth keeping after thirty days
Launch-week impressions are unreliable: every letter is at its best when it is new to you. After a month, the useful signals are quieter. You have acted on something from it — sent it to a colleague, tried a tool, changed a decision. You can predict what it will cover tomorrow and you still open it, which means the value is in the selection and the framing, not in novelty. The corrections, when they happen, are made visibly rather than silently. And the issue length tracks the news: short on a quiet day, long when something happened, never padded to hit a word count.
The reverse signals are just as clear. A letter you skim for the one link and delete is a feed, and a feed is better consumed as one. A letter whose "why it matters" line could be attached to any story is not adding judgement. And a letter that has grown a second and third daily send, each with its own sponsor, has changed its business without telling you. None of these mean the letter is bad for everyone; they mean it has stopped earning its place in your slot, and the slot is the scarce thing.
We apply the same tests to our own lanes, which is why every issue on this site is archived with its full text, its source list and its word count. If an Agent Signal lane stops passing them, the archive will show it before we do.
Questions people ask
Are AI newsletters free? Most of the dailies above are free and ad-supported. Several weeklies (Interconnects, Latent Space, Ahead of AI, ChinaTalk) run a free tier with paid extras. The Agent Signal lanes are paid: $9.99 a month per lane or $19.99 for all of them.
Which AI newsletter is best for beginners? The Rundown for news, The Batch for understanding, and our own The AI Shortcut if you want the beginner framing every day.
How many should I subscribe to? One daily and one weekly is enough for almost everyone. Add a vendor or specialist lane only if your job depends on that vendor.
Do you accept payment to be on this list? No. There are no sponsored placements here and no affiliate links. We will list our own letters last and label them, every time.
How often is this page updated? Facts are re-fetched and the title year checked at least every five months; the "last reviewed" date at the top is real. If a card is wrong, tell us and we will fix it and note the correction.
Try The Agent Signal · or read what an AI agent is and what agentic AI means first.
Get it in your inbox. latest AI news, AI agents, tips & tricks, in 5 minutes. Free.
Subscribe free