Summaries > AI > Ai > The Newsletter Is Dead. Hereâ€™s How To Survive AI Agents...

The Newsletter Is Dead. Hereâ€™S How To Survive Ai Agents

https://www.youtube.com/watch?v=b-1vvLrsyjE

TLDR AI is reshaping how creators write, edit, distribute, and monetize their work, while making human judgment and curation more valuable. Dan Shipper recommends experimenting with AI agents and personalized skills, measuring their performance, and separating prototypes from products that need support. As agents take over tasks like inbox triage and content creation, creators should build distinctive work, diversify their channels and revenue, and focus on reaching audiences in ways that genuinely matter to them.

Key Insights

Start by bringing AI into your existing creative workflow

Use an AI agent inside a familiar workspace, such as ChatGPT or Claude alongside Google Docs, rather than waiting for a perfect new tool. Ask it to research, critique a draft, reorganize sections, read passages aloud, or suggest revisions while you remain in control of the work. For substantial edits, have the agent record its changes as tracked changes so you can review and refine them. Treat the model like a capable collaborator, not an unattended replacement for your judgment.

Turn repeated feedback into a reusable skill

If you regularly give the same kinds of edits, collect examples and use them to create an AI skill that reflects your standards. A skill built from past copy edits, for example, can help writers bring drafts closer to an editor’s style before human review. Start by asking an AI tool to draft the skill, then try it on real work and note where it misses. Improve it with accepted, rejected, and newly added edits so the guidance becomes more useful over time.

Measure whether your AI skill is actually improving

Create an evaluation set of representative tasks to test a skill on the work it is meant to perform. Compare the AI’s outputs against your expectations, then feed the failures and missed cases back into the skill. Repeat the evaluation after each meaningful change instead of relying on a general impression that the tool seems better. This feedback loop makes progress visible and helps prevent a skill from becoming polished in theory but unreliable in practice.

Choose AI tools based on usefulness, not fear of missing out

Experiment with several platforms if you enjoy comparing tools, or focus deeply on one that fits your work. Switch when another tool genuinely improves your results or makes the process more enjoyable, not simply because it is receiving attention. Build workflows around concrete needs, such as recording where you met a contact, enriching their details, and prompting a follow-up. Keep the setup practical, and check what permissions or device requirements it depends on.

Use AI to expand your creative output without lowering your standards

AI can help turn books, notes, interviews, and existing research into new formats such as articles, podcast scripts, and videos. Use that capacity to pursue projects you genuinely want to exist, rather than producing more content just because creation has become easier. As output grows, prioritize editing, curation, and discernment so that quality keeps pace with volume. Distinctive, well-researched work remains important for earning attention in crowded feeds.

Adapt audience distribution beyond the traditional inbox

An email address is still useful, but it no longer guarantees that a reader will see a newsletter in a crowded or AI-managed inbox. Test multiple ways to reach your audience, such as a website feed, search, social channels, community platforms, or agent-friendly updates. Watch for meaningful engagement rather than treating subscriber counts or open rates as proof that people are paying attention. Build work and relationships strong enough that readers seek you out when they want more.

Separate rapid experimentation from products you promise to support

Give a small team room to test new models and build rough prototypes without making every experiment part of the core product. Label prototypes clearly as experimental, then evaluate whether they solve a real problem before committing to maintenance and customer support. Once an idea proves useful, refine it into a focused tool and assign clear ownership. This approach lets teams explore new possibilities while protecting the reliability of products people already depend on.

Diversify your business and keep developing human expertise

AI can change search traffic, distribution, and monetization, so creators should avoid relying on a single revenue source. Consider a mix of subscriptions, courses, sponsorships, consulting, and other offerings that fit your audience. At the same time, recognize that greater AI productivity can increase the need for people who provide judgment, editing, engineering, and coordination. Use AI to raise your ambitions, while continuing to build the expertise and trust that make your work valuable.

Questions & Answers

How can professional creators use AI agents in their writing workflow?

Work alongside an agent in tools such as ChatGPT or Claude Desktop: ask it to research, critique and reorganize drafts, read passages aloud, and apply document edits. Review and refine its suggestions as you would a human editor.

How did Every create and improve the “Kate pass” editing skill?

Every turned 30,000–40,000 of editor-in-chief Kate’s past copy edits into a skill writers can use to bring drafts closer to her style. The skill improves when Kate accepts, rejects, or adds edits.

How can someone build a useful AI skill and improve it over time?

Collect examples of the work, ask a model such as Claude, Codex, or ChatGPT to create a skill, and test it. Build an eval—a set of tasks that measures performance—then use missed cases and new examples to refine the skill and measure its performance again.

Should people use several AI platforms or focus on one?

It depends on the person. Experimenters may enjoy using several platforms, while others may prefer to focus deeply on one. Switch tools when it genuinely helps or is enjoyable, rather than from fear of missing out.

How should creators think about AI’s effects on their work and businesses?

The impact varies: AI is changing discovery, distribution, sales, and monetization, but some creators may benefit while others struggle. Diversify revenue across options such as subscriptions, courses, advertising, and consulting, and use AI to pursue work and opportunities that matter to you.

What new opportunities could AI create for apps and personalized feeds?

Individuals and communities will increasingly be able to build bespoke apps quickly without technical skills. AI agents could also assemble feeds tailored to a person’s needs and preferences—for example, flagging competing meeting priorities, reviewing job applicants, or surfacing relevant team experiments.

What is the next challenge as AI makes content creation easier?

Creation is becoming less of a bottleneck, while the volume of content is outpacing people’s ability to review it. The next challenge is improving AI’s editing, curation, and discernment capabilities.

Why will companies still need engineers as AI makes developers more productive?

Engineers will help ensure that AI-generated code and contributions work together safely and reliably. AI may also raise teams’ ambitions about what they can build, potentially increasing demand for engineering even as the role changes.

Is the newsletter dead, and how might AI change email?

Newsletters will not disappear, but an email address no longer guarantees that a person will see a message. AI assistants may prioritize important email, summarize the rest, and let users act on a small number of items. Creators will need to find ways to be noticed and prioritized by those assistants; open rates may become less meaningful.

How should newsletter writers prepare for AI agents reading email?

Test alternative ways to reach audiences, such as Slack alerts, website feeds, apps, SEO/AEO, and MCP integrations. Experiment with emerging channels now and learn which ones are worth investing in.

What should creators prioritize when using AI?

Use AI to make work you genuinely want to exist, rather than adopting it out of fear or letting it replace your craft. Create distinctive, high-quality work that stands out, and adapt as discovery and distribution change.

How does Every separate experimentation from building reliable products?

A frontier team explores new models and prototypes, while the product team builds and maintains stable customer products. Experiments are options, not obligations: product teams adopt those that prove useful and can leave the rest behind.

How should teams manage experimental software and its maintenance?

Clearly label experiments as unsupported, and provide ongoing support only after deciding to commit to an idea. This keeps exploratory work from being mistaken for a reliable product; agents may also make simple-app maintenance cheaper by helping triage and fix issues.

What is Every’s core business model?

Its core offering is a $30-per-month subscription that includes everything, supplemented by brand sponsorships, consulting, and courses.

What does the guest say about AI and the future of work?

AI and other technologies can eliminate jobs and disrupt livelihoods, but the future of work is more complex than a simple prediction of mass unemployment. The guest expects fears that nearly everyone will lose useful work to look naive, and argues that meaningful human work will remain.

Summary of Timestamps

Dan Shipper recommends that professional creators learn to work inside an AI agent harness such as ChatGPT or Claude Desktop. In his workflow, he can speak with an agent while writing, ask it to research or revise passages, hear text read aloud, and make document edits without stopping his own work. He highlights faster computer-use models and Gemini’s writing ability, including its help reorganizing an 8,000-word story about Codex.

The speaker describes using ChatGPT to coordinate Gemini’s critique and reordering of a document section, then apply the proposed revisions as Google Docs tracked changes. The edits were close enough to accept and refine, much like work with a human editor, before two human editors reviewed the piece.

Every created a “Kate pass” to help writers prepare drafts in the style of editor-in-chief Kate. The skill draws on roughly 30,000–40,000 of her past copy edits and improves as she accepts, rejects, or adds suggestions. This helps Kate extend her influence without personally editing every draft.

Building a genuinely useful AI skill is still difficult, though the speaker expects it to get easier over the coming months. For now, users can collect examples of their work, ask Claude, Codex, or ChatGPT to create a skill, and improve it by feeding in cases it misses.

The recommended way to improve a skill is to create an evaluation, or eval: a set of tasks that measures how well the model performs a particular kind of work. The model can then refine the skill to improve its eval results, using new examples and repeated testing to create a feedback loop.

The conversation turns to choosing AI platforms. The speaker says the choice is personal: curious experimenters may enjoy using several tools, while others may prefer to focus on one. Since major platforms may become more alike over time, people should switch because a tool is genuinely useful or enjoyable—not simply because they fear missing out.

One participant describes connecting a preferred AI tool to personal CRM tasks: recording where new contacts were met, enriching contact details, and prompting follow-ups. Similar workflows may work with tools such as Grokbot or Dots, though access to iMessage can depend on having a Mac available.

The speakers discuss anxiety about AI and the way new technology can initially seem shallow, impersonal, or threatening. One speaker is optimistic about technological change, comparing resistance to AI with historical reactions to the printing press, while emphasizing that appreciation for books and tradition can coexist with curiosity about new tools.

The speaker describes an “agency rupture” that can happen as people encounter AI. First, it performs work they thought was uniquely theirs; next, they see that it still needs supervision; finally, they realize that directing and refining its work is itself complex and creative. The unfamiliar may expand what people can do rather than erase everything they value.

AI is changing search traffic, book sales, distribution, and monetization, but not uniformly: some creators are struggling while others benefit. Tim Ferriss’s *The 4-Hour Workweek* may need updating for today, yet its idea of extreme leverage could be newly relevant. Creators are encouraged to diversify across subscriptions, courses, advertising, and consulting while exploring how AI can help them accomplish more.

The speakers expect individuals to build highly customized apps for themselves or their communities over a weekend, even without technical skills. Johnny Miller rebuilt his State Shift app in a weekend with Fable after an earlier technical partnership did not work out, giving him a useful product and a tighter feedback loop.

AI could make feeds more tailored and less dependent on the enormous audiences and constant user-generated activity that traditional social networks require. Agents might combine information based on both observed preferences and explicit requests—for example, flagging conflicting company meetings, reviewing applicants for a role, or surfacing experiments from a frontier team.

AI has made producing content much easier: creators can turn books, notes, and other source material into blog posts, podcast scripts, and videos. The new bottleneck is review. As creation outpaces people’s ability to assess it, AI’s editing, curation, and discernment need to improve alongside its generative abilities.

Agent swarms can produce a great deal of work, but users may not know when they are going off track—a mismatch between AI capability and human oversight. Every is building high-signal “frontier alerts” in Slack, where agents draw on users’ workflows and interests to surface relevant model developments. A larger goal is to learn from people’s reactions so AI messages reflect their taste without requiring them to inspect every output.

AI increases developer productivity, but the speakers do not expect companies to stop needing engineers. When more people can generate code or submit quick pull requests, engineers remain essential for making sure contributions work together safely. AI may also encourage teams to attempt more ambitious projects, increasing engineering demand even as the role changes.

The claim that “the newsletter’s dead” was deliberately provocative; the speaker still has around 120,000–130,000 subscribers and does not mean newsletters will disappear. Rather, the old assumption that an email address gives a creator ownership of a relationship is weakening as inbox tabs limit visibility and AI tools increasingly handle email.

The speakers expect email and SMTP to remain as communication infrastructure, while AI assistants increasingly decide what users should see, summarize less important messages, and help them act on a small number of items. For creators, being included and prioritized by those assistants may matter more than open rates. A meaningful audience consists of people who notice when a creator stops publishing and seek their work out.

Newsletter writers should prepare for AI agents to read more emails than people do, meaning an address no longer guarantees human attention. Every is testing other ways to reach audiences, including Slack alerts, a website feed, a possible app, SEO/AEO, and an MCP server that could notify ChatGPT about new articles. The aim is to learn which emerging channels merit investment.

Creators are encouraged to use AI to make work they genuinely want to exist, rather than adopting it from fear or allowing it to replace their craft. The speakers point to Every’s 8,000-word Kodak article, which combines interviews, custom animated design, and linked quotations that readers and agents can inspect. These agent-friendly formats could open new possibilities for articles, video, and games.

Despite shifting platforms and discovery methods, distinctive, high-quality content remains essential. The speakers discuss the writer’s success in building an engaged audience by covering an important topic with depth and care. Distribution tactics may change, but strong writing can continue to attract readers—and creators still need to adapt to how people find their work.

The conversation highlights how much AI tools can do beyond basic prompting. One example is an AI independently checking newsletter sites and DNS records to identify the platforms they use. The speakers encourage users to ask AI what they should do next; Every’s article “The Eight Levels of AI” describes moving from prompting toward coordinating teams of agents.

Every’s business combines a $30-per-month subscription with brand sponsorships, consulting, and courses. The host also promotes *The Ladders of Wealth*, available for pre-order. The discussion presents multiple revenue streams as a way to support the core subscription business.

Every separates experimentation from reliable product development. A frontier team explores new models and ideas, while the product team builds and maintains stable offerings. Experiments are treated as options rather than obligations: product teams can adopt what proves useful without making exploration everyone’s job or distracting people responsible for shipping.

Within the frontier team, “pirates” rapidly build rough prototypes, while “architects” refine promising ideas into focused tools that others can assess. Kit’s app store similarly allows people to build extensions without putting its core email platform at risk. Experiments are labeled unsupported in Labs; once the company commits to one, it takes responsibility for maintaining it.

The speakers discuss making software maintenance manageable by clearly separating experiments from supported products and deliberately deciding when an experiment becomes a commitment. At Kit, engineers may begin work with a pull request already waiting, which they can accept, modify, or replace. AI agents may also make simple apps cheaper to maintain by triaging bug reports, fixing issues, and responding to customers.

Dan’s closing hot takes are that Gemini is an excellent writing model and that fears of AI leaving almost everyone without meaningful work may eventually seem naive. Technology can displace workers, as the story of Andrew Carnegie’s father—a weaver affected by the loom—illustrates, but the future of work is more complicated than mass unemployment.

The guest shares recent reading: Antoine de Saint-Exupéry’s *Wind, Sand and Stars*, a memoir about flying mail routes in the Sahara, and Victor Hugo’s *Les Misérables*. He values balancing new ideas with classic works and recommends Every’s guide to “the eight levels of AI” and its essay “After Automation,” about why meaningful human work can remain after automation.

The conversation closes with places to find Dan Shipper, including X at @danshipper and Every.to. The host recommends episode 101 with J Singh for listeners interested in writing by talking to AI, noting that J helped automate Kit workflows with voice agents and built one live on the show. The host asks viewers to like and subscribe.