Week 44 · 2025-10-27 → 2025-11-02 · 23 newsletters

After SEO, Before Agents

the-post-seo-shift · ai-builder-identity · gtm-and-lifecycle · engineering-and-infra · essays-and-grace-notes

A sparse inbox week: 24 emails across seven days, with a few clear through-lines and a long tail of one-offs. The biggest cluster was marketing writers all circling the same observation from different angles, that the SEO era is ending and AI discovery is what comes next. The second was a small but unmistakable AI-builder identity conversation. After that the week thins out into engineering history, GTM tactics, and a handful of personal essays. No dominant news story.

The Post-SEO Shift: Marketers Naming the New Regime

Three writers, three angles, one regime change. Amanda Natividad at The Menu opened the week with "The Wipeout: What Comes After SEO," her framing of the post-SEO content world. The keyword-and-publish playbook is dead, she argues, because traffic sources have fractured, Google's share of attention has slipped, and AI answers questions before anyone clicks. Her prescription is platform-agnostic content built on a "source of truth" page on your own site, fed by feedback from every other touchpoint. The blog post is no longer the destination, it is the reference layer.

Kyle Poyar at Growth Unhinged ran the data version of the same argument twice in one week. Wednesday's piece was the 2025 State of B2B GTM report with Maja Voje, surveying 195 GTM leaders and finding that the average software company now runs five core channels plus another 5.5 in active experimentation. The takeaway buried in the numbers: distribution is fragmenting faster than teams can keep up. Sunday's follow-up, "Traffic is no longer a reliable growth metric", made the case more bluntly with Webflow as the proof point. 10% of Webflow signups now come from AI discovery, growing 4x year-on-year. 91% of those LLM referrals come from ChatGPT alone. ChatGPT traffic converts at 24%, which is 6x Google. Two-in-three LLM-referred conversions close within seven days.

Craig Zingerline at Growth Led ran the operator complement: five lifecycle email and SMS flows you should have running once acquisition cools. The argument is that once you reach early product-market fit, most growth comes from doing more with who you already acquired, not chasing more top-of-funnel. Read alongside Amanda and Kyle, the picture is consistent. Acquisition is getting weirder and the channel mix is getting more chaotic, so retention and lifecycle become the load-bearing parts of the funnel.

The take: this is the most coherent thread of the week, and it is worth taking seriously. Three independent writers, none coordinating, all describing the same shape: discovery is moving to AI, traffic is decoupling from conversion, and the playbook that worked for the last decade no longer maps. The Webflow numbers in Kyle's piece are the most concrete evidence anyone has put on the table this quarter that AI search is not a future trend. It is already a meaningful share of pipeline at one specific company, and there is no obvious reason that company is uniquely positioned.

AI Builder Identity: The Role Collapse

Two pieces this week wrestled with what AI is doing to product roles, from opposite ends of the seniority curve. Marily Nika at AI Product Academy opened the week with "The Dawn of the AI Builder," reporting from her conversations with CEOs and founders who have eliminated entire layers of PMs, merged design and product marketing into single roles, or pushed designers and engineers to "step in" where PMs used to sit. Her framing is stark: there will be people building with AI and people who are not, and the latter group will not be building at all. The polarized read is the point. She is not predicting it, she is reporting it.

Xinran Ma at Design With AI ran the practical counterpoint: a video tutorial on how product designers should actually use Claude Code, including Figma MCP setup and converting designs to code through that bridge. The implicit argument is the same as Marily's, just from the other direction. Designers are not waiting to be told what their job is now. They are picking up the tools.

The take: these two pieces sit on the same trend from opposite chairs. Marily is documenting what executives are doing to org charts. Xinran is documenting what individual contributors are doing in response. The interesting question for 2026 is which side moves first in any given company. The companies where executives restructure before ICs upskill are going to have a worse year than the companies where ICs upskill and executives notice.

Engineering, Infra, and the Long History

The week's most substantive engineering reads sat at opposite ends of the stack. Teiva Harsanyi at The Coder Cafe wrote up the Therac-25 case, the 1980s radiation therapy machine where a computer-controlled redesign removed an "independent safety net" of hardware interlocks and killed multiple patients. The lesson is not new but worth re-reading every couple of years, especially in a week when AI agents are being handed more autonomous authority. Software replacing hardware safeguards is a category of mistake that does not retire.

Kerman Kohli launched RouteMesh, an RPC routing layer for blockchain data, with a long history of the RPC market: Infura's early dominance when there were two chains that mattered, Alchemy and Quicknode's rise on the back of CryptoKitties-era data growth and the multi-chain explosion, and his case that the next phase is a grid layer abstracting providers the way electricity markets eventually abstracted producers. Whether or not you care about crypto infrastructure, the framing of "vertically integrated to grid-abstracted" is a pattern worth keeping in your back pocket.

Addy Osmani at Elevate wrote the cleanest current statement of where AI coding is going with "Conductors to Orchestrators: The Future of Agentic Coding." Up to 90% of software engineers now use AI for coding, he writes, but a paradigm shift is underway from working closely with a single agent (Conductor) to coordinating fleets of agents in parallel without interference (Orchestrator). The role of senior engineers is shifting from "how do I code this" to "how do I get the right code built." Read it next to the Marily piece on PM role collapse and the symmetry is the story.

Ben at SeattleDataGuy hit his 200th newsletter with a meditation on how the data world spans multiple decades simultaneously: teams running Netflix-grade infrastructure sitting next to teams still living in 2014. Internal Tech Emails republished a 2004 Jim Allchin internal rant to Bill Gates and Steve Ballmer titled "losing our way," in which a senior Microsoft exec admits he would buy a Mac if he were not a Microsoft employee. It reads as a historical artifact and as a permanent template for what happens when a company forgets its customer.

The take: the engineering writing this week was unusually coherent on one point. Whether the topic is medical devices in the 1980s, RPC infrastructure, agentic coding, or Microsoft circa 2004, the through-line is that abstraction layers move trade-offs around but never eliminate them. The Therac-25 lost an interlock. The RPC market lost vendor diversity. The Orchestrator paradigm trades synchronous human judgment for parallel agent throughput. Microsoft in 2004 traded customer focus for feature breadth. Worth holding in mind as the agent stack matures.

Essays and Grace Notes

A handful of pieces that did not fit the main threads but deserve names. George Milton at Gross to Net wrote "Founder Beware: Don't Take That Investor Check Before You Read This," arguing that PE and VC optimize for maximum wealth in a 7-to-10 year window while CPG founders optimize for passion, meaning, and brand, and that these are structurally incompatible no matter how transparent everyone is upfront. It is the cleanest one-page statement of the founder-investor misalignment problem I have read this year.

dynomight at Dynomight ran "Dating: A mysterious constellation of facts," puzzling through why dating apps are hated, popular, and technologically trivial to build, yet no challenger displaces the incumbents. Match Group's 25% operating margin (Apple-tier) is the load-bearing data point. Steven Schlafman at Where the Road Bends wrote "Stunned into Consciousness," a long personal essay on a near-overdose at a Roots concert during Super Bowl weekend that became a turning point in his recovery. Ben Kassoy at A Strawberry Spinning Like A Dreidel recapped a residency at Emory, his alma mater. Justin Mares at The Next ran a long product-recommendations piece on tools and supplements for healthier living, originally written as a guest post for Lenny Rachitsky.

Rob Thomas at The Mentor hit on Kidlin's Law, the idea that writing down a problem clearly is half the work of solving it. Medha Nair at Fincredible ran a deep dive on the jewelry retail value chain. Matthew Klein at The Overshoot wrote "Inside America's Consumption Recovery," arguing that fears of a K-shaped consumption boom dependent on rich consumers and rising stocks are misplaced. Real PCE rose 4% annualized from May through August. The data does not match the doom narrative.

Snaxshot ran a sober note that 42 million Americans, including 16 million children, were set to lose SNAP benefits on November 1, calling it "the greatest hunger catastrophe since the Great Depression." Method Studio compiled creative-director clips on how to use decks to sell ideas. Javier Lopez at VC Hiring ran the weekly job board. Upen at Micro SaaS Idea ran the $1K-$10K MRR product roundup, with Rosie (an AI answering service) crossing $1M ARR as the headline number. Richard King at The Product Marketing Drop wrapped the week with product marketing horror stories from the Boston PMM Summit.

Three Takeaways from the Week

The post-SEO shift went from theory to numbers this week. Amanda Natividad named the regime, Kyle Poyar put data on it, and Craig Zingerline ran the operator response. If you work in growth or marketing and you do not have a serious answer to "what is our AEO strategy" by end of Q4, you are behind. Webflow's 24% conversion rate on ChatGPT traffic is the line that should land in your next planning meeting.

The role-collapse conversation is now coherent enough to plan around. Marily Nika is documenting executives restructuring product orgs, Xinran Ma is documenting designers picking up Claude Code, Addy Osmani is documenting senior engineers moving from coders to orchestrators. The trend is the same trend, viewed from different chairs. The companies that survive the next eighteen months will be the ones where executives and ICs are upskilling at roughly the same pace. The mismatches will get expensive.

If you only revisit three pieces from the week, I would suggest Kyle Poyar on traffic no longer being a reliable growth metric for the cleanest data point on AI discovery, Addy Osmani on conductors to orchestrators for the cleanest current frame on agentic coding, and George Milton on why founders should think twice before taking the investor check for the cleanest one-page statement of structural misalignment between CPG founders and their funders. Three pieces, three different problems, one shared instinct: name the new shape before you act on it.