AI Newsletters for Marketers: The 10 Worth Reading (2026)

AI marketing newsletters compared

An AI marketing newsletter is a recurring email that keeps marketers current on AI and on its use in their work: new assistants, ad tools, ways of writing prompts, the way AI answers now send visitors to websites. The first six below are written for people in marketing jobs. The last four are general AI dailies, daily newsletters for any reader.

Every time here counts all the words of one issue, ads included, so you can compare the 10 on one measure. Skimming takes less, and opening the links takes more.

The speed used is 238 words a minute. It is the average for silent reading of English non-fiction, from a 2019 review of 190 studies by Marc Brysbaert, a psychologist at Ghent University.

The counts cover the five weeks from Monday 28 September to Sunday 1 November 2026. They come from each newsletter's public archive, the page that lists its past issues. This Week in AI and The AI Marketer Weekly keep no public archive, so their entries rely on what their publishers show. Prices were read on 4 November 2026.

This site also sends its own newsletter, for people building a business alone with AI, and leaves it out of this list because a site shouldn't rank itself.

Which marketing AI newsletter to read first

The table gives the facts, and what to read first depends on your job. One pair fits most readers, and you can check it before you subscribe, since both keep a public archive.

Some jobs add to the pair or change it. Most lines point to something you can open before subscribing.

• You do all the marketing alone. Keep the pair above: the week's news and one longer read, in about 25 minutes. Add a newsletter from a line below only if that job fills most of your week.
• You write content or run social media. Keep the pair above. Marketing Standup's issue of 2 November 2026 covers Instagram's new format for long videos. Almost Timely's essay of 4 October shows how to save your instructions for AI agents, programs that carry out tasks on their own, so you stop rewriting them.
• You want explainers on AI in content work. Add The AI Marketer Weekly, a digest of articles by Brafton, a content marketing agency. Brafton's roundup of AI marketing news for October 2026 shows the kind of article it lists.
• Ads take most of your marketing budget. Add Aleric Heck's The AI Marketing Newsletter. His morning edition of 29 October shows how to have an AI assistant judge your ads and video thumbnails, and his evening emails carry ad ideas.
• You lead a marketing team. Read This Week in AI, from Marketing AI Institute. Add its podcast, The Artificial Intelligence Show, if you commute: about 90 minutes on Tuesdays, with other episodes on some Thursdays. In the one of 24 September, the marketing chief of Baptist Health, a Florida healthcare group, tells how her team took the lead on AI.
• You also build your own product with AI. Add Ben's Bites, a newsletter for people who build that way. Its writer's post of 29 October asks how much of the AI news a builder needs.
• Following the day's AI news is part of your job. Add one daily. The Rundown AI has carried the most marketing how-tos lately. Superhuman AI comes every day in about 5 minutes. The Neuron reads lighter and can come once a week. TLDR AI suits technical readers.

AI newsletters for marketers

These six are written for people in marketing jobs, which is why they come before the general dailies.

Marketing Standup, from Superhuman AI

Marketing Standup is the weekly marketing edition of Superhuman AI, one of the largest AI dailies. It promises the week's marketing news in 5 minutes. Its five issues from 28 September to 26 October 2026 ran 1,175 to 1,300 words, about 5 minutes each. Its site says "115k+ marketers and CMOs" read it, CMOs being chief marketing officers.

All its issues are signed "Zain, Faiq, & the Superhuman AI team". Zain Kahn is a founder of Superhuman AI, and Faiq a colleague. The issues say the team spends "over 100 hours each week curating". They don't say whether AI helps write them.

The issue of 2 November shows what you get. It opened on DoorDash's $17 billion deal to buy Instacart, which could create a new advertising network for groceries and deliveries. Prada's holiday ad came next, made with veteran Disney artists and praised for its human touch. The third story was Krea's Ad Studio, an AI service that turns a company's web address into up to 12 static ads.

The rest of each issue follows one plan. It gives the week's campaign hits and misses, then tactics from other people's posts and videos, then one longer piece, then new marketing apps. On 26 October the longer piece covered software companies adding comedy writers and stunt producers to their marketing teams.

AI is one angle among others. In most issues from 28 September to 2 November, one of the three top stories was about an AI product. The issue of 12 October had Meta's assistant for advertisers. The issue of 6 October had an ad generator from Higgsfield, an AI video company.

The issue of 19 October covered new features for finding customers in Lovable, a service that builds apps from a description. One issue, on 28 September, walked through a task: eight steps with Runway, an AI video service, to turn a product photo into a video for a product review.

Marketing Standup comes first because it is the only news email here that is weekly with a public archive you can check. It gives less method than an essay, which is the reason to add one.

Almost Timely News, by Christopher Penn

Christopher Penn, co-founder of the marketing analytics firm Trust Insights, writes Almost Timely every Sunday. It was free on 4 November 2026, and Substack, the platform that hosts it, showed over 293,000 subscribers. A long weekly piece takes longer than a digest, but it comes from a person who tests things and reports what happened.

His five essays from 4 October to 1 November 2026 ran 2,817 to 3,552 words, 12 to 15 minutes. The whole issue takes 18 to 26 minutes once you count ads for his courses, job listings from his online community and his event dates.

Each issue links a video and an MP3 version. Each also opens with a "Content Authenticity Statement" saying how much he wrote himself. On 1 November it read "100% of this week's newsletter was made by me, the human". On 18 October it read 95%, because ChatGPT made the lead image.

The essay of 18 October explains open-weights models, the AI models anyone can download and run on their own machine. The one of 25 October lists the AI apps he uses.

The one of 1 November explains what "agentic" means. His short answer is "Agents have agency": once you hand over a task, the agent does it, and you only review the result. He then sorts AI products into five levels, by how much they do without you. OpenClaw and Hermes Agent, two agents you run yourself, sit at level 4, "Done without you".

The same essay walks through a whole task: an agent that finds new customers. He builds it from three sets of saved instructions. One searches the web, one checks whether each company fits, and one drafts the first approach. A person approves each draft before it goes out.

A smaller example takes a minute to try. "What's Changed in Basic AI Prompting" (20 September 2026) argues that today's models are built to carry out tasks. So you give them the facts and say what a finished result looks like, his "definition of done". He changed one line in his own prompts:

Before: "Ask me one question at a time until you have enough information to succeed at the task."
After:  "Ask me up to 20 clarifying questions until you have enough information to succeed at the task."

With no limit on questions, he writes, models ask three to five and guess the rest. With a limit, they ask about 11 on average. He also calls two phrases people used to add to prompts obsolete: "think this through step by step" and "take a deep breath". These are his own observations, not a published test, so try the new line in your own assistant and count the questions.

Almost Timely suits anyone who wants to learn how to work with AI rather than what launched this week. If you only have 12 to 15 minutes, the essay is the part to read.

Growth Memo, by Kevin Indig

Kevin Indig advises companies on organic growth, the visits that come from search results without paying for ads. He writes Growth Memo "for people curious about the intersection of marketing and business strategy". He writes from the angle of search. Substack shows over 30,000 subscribers.

AI search here means the answers that ChatGPT, Google and similar assistants write for a question, and how brands show up in them.

His essays show the numbers behind each conclusion. The free essay of 19 October asked whether visits from AI answers hide in the "Direct" line of Google Analytics, where it puts visits it can't trace to a source. He looked at 12 sites from January to September, counting all their Direct visits, then the engaged ones alone. An engaged visit lasts longer than ten seconds, covers two pages or more, or ends in a sale or a sign-up.

All Direct visits rose 140.9%, while engaged ones changed by −1.0%. So the growth was mostly bots, automated programs and not people: one site logged 409,000 Direct visits, of which 198 were engaged. His answer to the question: he "could not yet detect" hidden AI traffic.

The free essay of 2 November asked which channels grow while search shrinks. It used Similarweb, a company that estimates websites' traffic, for 10,000 sites in the US. Visits from email grew 16.3% in the first eight months of 2026, against the same months of 2025.

The free essays come on Mondays and take 7 to 9 minutes (1,581 to 2,168 words). Paying subscribers also get the Growth Intelligence Brief, a report on which sites gain or lose visibility on Google, and resources such as a planning workbook. A subscription cost $15 a month or $150 a year on 4 November 2026.

Each essay stays free for about four weeks. On 4 November, the essay of 5 October had already turned paid-only, while the one of 12 October was still open to all.

Growth Memo suits you if work on Google and AI answers fills most of your marketing time. Best AI Visibility Tools compares twelve trackers of how often a brand appears in AI answers.

This Week in AI, from Marketing AI Institute

Marketing AI Institute is an education, events and media company for marketers. Paul Roetzer, who leads it, also founded SmarterX, an AI education company for businesses. The institute's weekly email, This Week in AI, promises "in-depth analysis of the top stories in AI each and every week". Its page says "40,000+ marketers and business leaders" read it.

No back issue is public. The only sample, on the sign-up page, is a screenshot taken on 7 November 2023, when the email was still called The Marketing AI Newsletter. It shows three top stories in one or two lines each, with a link, then "Other News to Watch". Judge the email by the first few issues that arrive.

What you can judge today is the team's podcast, The Artificial Intelligence Show. Roetzer and Mike Kaput, SmarterX's chief content officer, go through the week's news on Tuesdays, in episodes of about an hour and a half.

The episode of 3 November covered the stock-market filing of Anthropic, the company behind the assistant Claude. It then turned to AI and the meaning of work, and to Microsoft's buyout offer to 60,000 employees. That is general AI news, discussed for business leaders.

The Thursday episodes are more specific. Some tell one company's case, such as how Baptist Health's marketing team built its AI training, on 24 September. Others answer listeners, such as "How do you keep up with AI without burning out?", recorded on 15 October at MAICON, the institute's conference.

Katie Robbert, chief executive of the analytics firm Trust Insights, said in January 2025 that she relies on the institute. Her reason is that it presents "here's what's happening versus here's what you should buy". This Week in AI suits marketing leaders who must decide what their team does with AI.

The AI Marketing Newsletter, by Aleric Heck

Its name sounds generic, but it is one person's newsletter. Aleric Heck writes it on beehiiv, a newsletter platform. He is founder and chief executive of AdOutreach, an agency that runs YouTube ads. He is also behind AdsCreator and ThumbnailCreator, two apps for making video ads and video thumbnails.

It sent 60 emails from 28 September to 1 November 2026: a morning edition every day, weekends included, and an evening email on weekdays.

A morning edition runs about 2,570 words, 11 minutes, on one AI story told for people who run marketing. The one of 4 November started from an Adobe study of more than a trillion visits to US retail sites. Shoppers arriving from an AI assistant brought 83% more revenue per visit than other shoppers in September.

The edition then explained how a store can get found by AI assistants. The rest of the email holds apps to try, a tip of the day, a sponsor's ad and ads for Heck's own apps.

The ad ideas are mostly in the evening email, which takes 3 to 4 minutes. Each one presents one of Heck's businesses, and many open with a usable idea first. The one of 16 October gave a three-question test for an ad's opening line. The one of 3 November matched a video's visual style to the angle of the ad.

It is the only daily here written for marketers, and it suits people who buy ads. The mornings alone take about 75 minutes a week, and the whole newsletter about 90.

The AI Marketer Weekly, from Brafton

Brafton, a content marketing agency based in Boston, launched The AI Marketer Weekly as a companion to its AI blog at contentmarketing.ai. The sign-up page promises a weekly email of its latest articles on AI in marketing, every Monday morning. The blog's latest articles on 4 November included a roundup of October's AI marketing news and a piece on whether links in AI search help a site's traffic.

No issue of this email is public, so judge it by the first few that arrive. Brafton's main newsletter gives an idea of the format, because Sequenzy, an email software company, subscribes to brands' emails and publishes what it receives. Its copies of Brafton's issues of 13 and 20 October list the agency's own articles, each with its reading time, then three links to marketing news sites.

By its sign-up page's description, The AI Marketer Weekly lists articles from the AI blog in the same way. It suits people in content work who want explainers more than news.

Daily AI news, and how much of it is marketing

The dailies are what most lists put first. They do one thing well: tell a wide audience what launched today. Their own words show who they write for. Pete Huang, co-founder of The Neuron, said in 2024, before its sale, that its aim was to teach "business leaders and knowledge workers" to put AI to work. TLDR AI's page says it is for "engineers, researchers, and technical professionals".

To see how much a marketer gets from them, every weekday issue from 28 September to 30 October 2026 was sorted by its headline and its how-to line. An issue counts when its lead story or how-to is about marketing, advertising or brand work.

So marketing is a small share of a daily. At The Rundown AI it was rare until late October: 1 issue in the first four weeks. Four of its eight how-tos from 26 October to 4 November 2026 were marketing tasks, the latest turning one product photo into three studio shots. That is too recent to count on. Superhuman AI's how-to of 3 November built a daily report on what people say about a brand on X.

The dailies also overlap. All four reported the same launches, such as the AI model Kimi K3.1 Pro on 26 October and Google's Genie 4 on 28 October. One daily gives you the news, and a second mostly repeats it.

The Rundown AI. Rowan Cheung started it after his Twitter thread about ChatGPT went viral in ChatGPT's first days. It arrives in the early morning, US time, on weekdays, with an interview edition on some Sundays. An issue holds a lead story, a piece by its staff and an "AI Training" how-to. It also holds two sponsors' ads, short news items and one reader's way of using AI, about 1,660 words in all.

Superhuman AI. Zain Kahn's daily promises AI "in just 3 minutes a day" and comes every day of the week, with a robotics special on Saturdays and a Sunday special. Its issues ran about 1,200 words, 5 minutes. The one of 3 November held three news items, a how-to in numbered steps, trending posts and prompts.

The Neuron. Noah Edelman and Pete Huang founded it in 2023, and TechnologyAdvice, the company that now sells its ads, bought it in January 2025. Its tone is lighter: its issues open with "Welcome, humans."

The Sunday issue of 1 November led on publishers selling ads to AI agents: a brand pays to place its message in a version of a page that only AI assistants read. A link at the bottom of each issue switches it to once a week, for readers who find a daily too much.

TLDR AI. TLDR, a publisher of short tech newsletters, sends it on weekdays to "1,100,000 subscribers". It gives short summaries of model releases, research and developer news, about 830 words in all. Each summary gives the length of the linked read, such as "64 minute read" for one essay on 30 October.

Two more general AI newsletters, outside the 10, appear on many lists.

Mindstream. Matt Village and Adam Biddlecombe launched it in May 2023 with five subscribers. Seventeen months later its platform, beehiiv, counted 210,000 "daily readers", and HubSpot, the marketing software company, bought it for HubSpot's media network. That network already held the business newsletter The Hustle. Mindstream's sign-up page promises AI and tech news "every morning" and states that HubSpot will use your details to contact you about its products.

On its website, each weekday is split into two short stories of about 220 words, 1 minute each, of general AI and tech news. On 4 November one story covered a warning from Anthropic: its models could soon help build stronger versions of themselves. Mindstream suits a reader who wants the day's AI news in two minutes and doesn't mind hearing from HubSpot.

Ben's Bites. Ben Tossell started it in 2022, six weeks before ChatGPT came out, while he worked on the AI team of Zapier, a service that links apps together. It is written for people who build with AI, and the team tries much of what it covers. Its AI specialist, Keshav, tried Google's Gemini 4 Argon on 15 October to edit a short video. He found it "great analysing reels and planning cuts", but its final edit had errors.

It came on Tuesdays and Thursdays from 28 September to 1 November: ten free posts of 417 to 1,416 words, typically about 3 minutes. That rhythm may change: on 3 November Tossell wrote that until the end of the year "the formats will vary, the content, even the days + frequency".

A community of builders cost $80 a year on 4 November 2026, and Substack showed over 118,000 subscribers. Ben's Bites suits a founder who builds the product and also does the marketing.

Other AI marketing newsletters in Google's results

Google's first page for "ai marketing newsletter" and its close variants also shows names outside the 10. On 4 November 2026, this is what they were.

• A LinkedIn newsletter has the same name as Aleric Heck's. Its entry on Google promises "curated news and insights" every week, without naming a writer. Subscribing goes through LinkedIn.
• AI Marketing is Anand Karasi's newsletter on Substack. Substack showed over 70,000 subscribers. Its "Best Read" posts gather figures from recent studies, and the latest was dated 2 October 2026.
• AI in Marketing is a free edition of AI Weekly. Its page lists it as weekly, for marketing teams. The latest issue it showed was dated 21 June 2026.
• The Cognitiv Edge comes from an advertising company. Cognitiv sells ad targeting that uses AI, and its newsletter promises "the latest news and views shaping AI advertising".
• Zapier's article lists AI newsletters in general. Four of its nine picks are on this page: Superhuman AI, The Rundown AI, The Neuron and Ben's Bites. The others include newsletters on AI research, policy and technology.
• Jasper's Newsletter Agent isn't a newsletter to read. Jasper sells marketing software, and this agent drafts a company's own newsletters.

What a subscriber count tells you

Every daily above prints a large number on its sign-up page, and a large list doesn't say how many people read it. The founders' own accounts explain where those numbers come from.

Rowan Cheung told the Creator Spotlight podcast in October 2026 that The Rundown AI has 2 million subscribers on its main daily, and over 3 million across its newsletters. He puts its open rate, the share of recipients whose mail app reports the email as opened, above 55%. That figure comes after removing inactive subscribers, which he says it does "very aggressively".

Ads on Facebook and Instagram are, in his words, "the bread and butter" of its growth. The sign-up itself also earns money. SparkLoop, a referral service, shows each new subscriber a list of other newsletters and pays the newsletter that showed the list for every sign-up. He says The Rundown AI made $30,000 to $40,000 a month this way before it ran any ads. So some subscribers on a list joined with one tick on another newsletter's page.

The Neuron started differently. Its co-founder Noah Edelman said its first 100,000 subscribers came from the LinkedIn following of his co-founder, Pete Huang. The money it then earned from selling ads went "basically all" into buying ads on Facebook and Instagram. Mindstream's founders also paid for growth, and found that "not all subscribers are equal": LinkedIn brought strong early subscribers, then Facebook and Instagram ads brought some of their most engaged paid ones.

What a bought subscriber is worth depends on the price. Nataasha Hallaldeen ran ad campaigns on Facebook and Instagram for newsletter publishers at the consultancy Inbox Collective. She wrote in October 2025 that very cheap sign-ups usually came from the wrong audience: "they'd rarely engage" and unsubscribed quickly. She settled on $1 to $2 per sign-up as the range that brought subscribers who stay and keep opening.

SparkLoop gives one client's case, with no date: it took five sign-ups from such ads to find one subscriber who stayed active. So a subscriber from a paid ad may read every issue or none, and the count doesn't separate them.

A writer can also choose a smaller list. Ben Tossell wrote on 29 October 2026 that he had just unsubscribed more than 60,000 people from Ben's Bites because they weren't opening it. A longer list would look better to advertisers, he admits: "the numbers are good for sponsors. It's just not the point of this whole thing."

Open rates are no proof of reading either. Apple's Mail Privacy Protection, a setting in the Mail app on iPhone and Mac, "prevents senders from seeing if you've opened the email message they sent you". With it on, Apple loads each email ahead of time, including the tiny image senders use to count an open, even if you never open the email. The sender's software can then count an open that never happened.

Litmus, a company that measures which mail apps people use, says this affects about 55 to 60% of all opens.

A list's size doesn't say who owns the newsletter, either. HubSpot owns Mindstream, and TechnologyAdvice owns The Neuron.

HubSpot has a reason to want newsletters. Kevin Indig explained it on SYSTM, a show on YouTube, in July 2026. HubSpot lost about 80% of its search traffic, he said, and the loss didn't show in its share price. His explanation is that it had already bought The Hustle and built "an army of newsletters, podcast, YouTube channels". A newsletter gives its owner an audience that doesn't depend on Google.

The Rundown AI, for its part, sells AI courses. Its sign-up quiz sends a marketer who is new to AI, in Cheung's words, to "that exact product right after you finish this quiz". So a free newsletter can also be its owner's way to reach buyers.

Who writes a newsletter tells you more than how many receive it. Cheung says The Rundown AI uses AI "for the structuring, for some parts of research, but we never do it for the human part". Penn states in each issue how much he wrote himself. A named writer who says how AI is used is easier to judge.

At the far end, Nieman Lab, Harvard's journalism lab, reported in January 2025 on a network of AI-written local newsletters in 355 US towns. Their sign-up pages shared the same seven reader testimonials.

Four checks before you subscribe

Best-of articles age quickly, and a newsletter can change after it is listed. Four checks take a few minutes in a newsletter's public archive: add /archive to its address. This Week in AI and The AI Marketer Weekly have none, so judge them by the first issues that arrive.

1. The date of the last issue. If nothing came out in the last month, skip it. monday.com, a work software company, calls Not A Bot a weekly in its January 2026 list. Not A Bot's archive holds three posts from December 2022 and nothing since.
2. What it really covers. The same list describes AI Breakfast as weekly deep dives for marketers. AI Breakfast sent 15 general AI news emails from 28 September to 1 November 2026, three a week.
3. A named writer, and what else they sell. Look for a signature, an about page or a byline. Then read the sponsors and the house products in two or three issues.
4. What is free, and for how long. Each Growth Memo essay turns paid after about four weeks, and Ben's Bites has a paid community. Read the paid offer before you count on the archive.

How many to read, by the time you have

Time is the real price of a free newsletter. SmarterX, the AI education company Paul Roetzer founded, surveyed more than 2,100 professionals in 2026, about a third of them marketers. It asked what keeps people from using AI. Lack of training came first (38%), then lack of understanding (35%) and lack of time (30%). The survey was promoted mostly through SmarterX's and the institute's own channels, so its sample leans toward people already using AI.

The same survey asked what people want to learn. 58% named how to fit AI tools into the work they already do, 51% AI agents, and 45% building assistants without writing code. Only 15% asked for prompting tips. That favors newsletters that show a whole task done with AI over lists of new launches.

Of the picks, Almost Timely most often shows a whole task, as with the customer-finding agent in its essay of 1 November. Growth Memo shows how to measure what AI does to your traffic, and Marketing Standup is there for the news.

The Neuron's weekly version gives a daily's news in less time, through the link at the bottom of each issue.

Some readers need less than any of this. Katie Robbert of Trust Insights puts the question differently: "it's not how do I keep up, it's do I need to keep up". In that January 2025 episode of In-Ear Insights, the podcast she hosts with Penn, the hosts weigh what a new model brings. For a content marketer writing blog posts, they put the gain in performance at "maybe like a two or three percent difference".

Newsletter writers feel the overload too. Ben Tossell began Ben's Bites to keep up with everything and sent it five days a week. He cut it to two days because "It was so noisy", and his revenue fell with it, he wrote on 29 October 2026. His advice to readers that day: "You don't need to try every tool under the sun."

Whatever you keep, read with one question in mind. Penn sorts every piece of AI news first: "is this a new model, a new harness, or a new application?"

• A new model. This is the AI itself, such as GPT, Claude or Gemini.
• A new harness. This is the software around a model that lets it act, from a chat app to tools that carry out whole tasks.
• A new application. This is a job you point the AI at, and these are the items most marketers can act on.

His routine for someone with 30 minutes a week: "Extract what matters from one or two good sources. Transform it through your own professional lens", then "Try one new thing a week". Marketing Standup and Almost Timely's essay take about 20 minutes, which leaves ten for that one thing.

A simple way to hold a newsletter to that one new thing is to write down, after each issue, what you will try this week. Drop any essay or how-to newsletter that gave you nothing to try in four weekly issues, or in two weeks of a daily. Four is a rule of thumb, not a measured threshold. Best AI for Marketing starts one step earlier, for readers who haven't settled on an AI assistant yet.

Build your own AI briefing instead

If your questions are narrower than what any newsletter covers, you can build your own briefing. Christopher Penn writes about AI for marketers every week, yet his issue of 8 March 2026 on how he keeps up lists no newsletter among his sources.

Almost every big AI company runs a community on Discord, a group chat app, with a channel for its announcements. Discord lets anyone copy such a channel into a group of their own. Penn follows them all in one place, so every launch lands there.

He adds Reddit communities where people test new models, conference talks on YouTube and research papers from arXiv, the site where researchers post them. He also follows the AI companies' developer blogs through RSS, a format that lets one app gather the new articles of many sites. Monitoring software brings in "five hundred articles or so an hour". A small AI model then scores each item for relevance before he reads the top ones.

A solo marketer can run a smaller version in the 30 minutes a week of Penn's routine:

1. Write down the two or three questions your work raises this quarter. For example: "Do AI answers send us visitors?" or "Which ad platforms now make the creative for us?". The advice on Trust Insights' podcast is the same: pick the part of your work that needs improving, then "go see what's going on in that particular slice".
2. Collect five to ten sources in one feed reader. A feed reader is an app that gathers the new articles of the sites you choose. NetNewsWire, for Mac, iPhone and iPad, was free on 4 November 2026. Paste each source's feed address into the app: openai.com/news/rss.xml for OpenAI, blog.google/technology/ai/rss/ for Google. Every Substack newsletter has one at /feed, such as almosttimely.substack.com/feed. A newsletter with no feed stays in your inbox.
3. Once a week, give your AI assistant the headlines closest to your questions. Copy about 20 of them from the feed reader into this prompt, which does the sorting Penn's small model does:

Here are this week's AI headlines: [paste the headlines].
My job: [your role and sector].
My questions this quarter: [question 1], [question 2].
1. Label each item: new model, new harness (the software around a model), or new application.
2. Score each from 1 to 5 for how much it changes my answer to one of my questions.
3. List the three highest scores, each with one thing I could try this week.

4. Read the three it picks in full and try one thing.
5. Drop the sources that gave you nothing to try in four weeks. It is the same rule as for newsletters.

Penn also tested the opposite: having AI write the newsletter. His issue of 30 August 2026 built a test newsletter written entirely by AI, from research papers to a 4,000-word issue in his own style. One issue cost 9 US cents, about $33 a year for a daily. That issue gives his prompts in order, for anyone who came here to write a marketing newsletter with AI.

His conclusion: "if you use AI to automate your outputs, you risk becoming weaker. If you use AI to automate your inputs - like this custom-built newsletter to teach me things - it can help you become stronger."

That is the case for a briefing of your own. Today's models "all pretty much know the same things", in Penn's words, and the dailies summarize the same launches. A briefing built from your own sources, your sector and your own questions is one nobody else receives.