Week 41 · 2025-10-06 → 2025-10-12 · 18 newsletters

Build Fast, Aim Slower

ai-labor-and-output · ai-search-replaces-google · writing-in-the-age-of-llms · macro-cracks · voice-and-reflection

A genuinely sparse week. Eighteen emails across seven days, no breaking news story, no dominant macro thread, and a noticeable number of writers either announcing personal transitions or asking quiet questions about where the AI buildout is actually pointed. The through-lines that did emerge were ones the inbox kept circling back to from different angles: AI's effect on team size, AI's effect on search, and AI's effect on the writing economy that most of these newsletters live in.

AI and the Shrinking Team: Two Operators per Magnitude

The most concrete piece of the week was Kerman Kohli's "The Machine Economy", in which he disclosed that RouteMesh went from seven people to two and that the consequences were faster shipping, lower burn, and more decisive movement on new ideas. His frame is that the link between economic output and headcount, which has held since the industrial revolution, is being severed by agents that scale one person's labor by an order of magnitude. He treats "five million dollars per person" as a normal top-decile number now. The argument is not new, but the willingness to publish actual headcount changes alongside it is.

Carilu Dietrich at Hypergrowth Leadership ran the production-side companion: Deel's AI Workforce launch video, which was itself generated by six creatives using six AI tools across six days. The detail that landed was that the team did not try to do it all in one tool. Whisk, Flow, Visual Electric, ElevenLabs, ChatGPT, and Gemini each owned a step. The piece's takeaway, that the most impactful uses of AI are not shortcuts but experts pushing the tools to their limits, is the right corrective to the demo-reel narrative.

The take: the operator-level story this week is that two skilled people with a stack of tools are starting to beat ten people with a process. Kohli's RouteMesh numbers and Dietrich's Deel breakdown are the same story told from the founder seat and the marketing seat. The question Kohli does not answer, which is what happens to the other five who got cut, is the question the rest of the year will have to answer for him.

Search, Rewritten: AEO Is the New SEO

Two newsletters converged on the same observation from different sides of the table. Kyle Poyar at Growth Unhinged profiled Docebo, a publicly traded LMS, which now sources 12.7% of its high-intent leads from AI discovery, up 429% year-over-year, with one team member running both SEO and answer engine optimization. ChatGPT is at 800 million weekly active users, up from 500 million in March. The four tactical takeaways Poyar pulled out of the Docebo case (self-reported attribution, branded search traffic, sharing schema, and pattern-matching what the LLMs actually cite) are the cleanest playbook published this week.

Max Mitcham at From the Ground Up ran the panic version of the same argument. His framing: when a buyer asks ChatGPT for the best social listening tool, they get a list of five to ten companies and no second page. You are either in the answer or you do not exist. He built an N8N agent system to monitor whether Trigify shows up across the major LLMs, and the unsettling part of the post is how few B2B companies have any visibility into whether they are being recommended.

Nesrine Changuel at Product Delight Tips ran the design-side counterweight, using Google Photos' new Highlight Video feature as an example of AI done as quiet enhancement rather than headline feature. Her thesis, "enhancement over dominance," is that the AI implementations that will hold up are the ones that remove friction without announcing themselves.

The take: the SEO-to-AEO transition is not a future scenario, it is already cannibalizing inbound funnels at companies measuring it. The B2B businesses still optimizing for SERP page one in October 2025 are competing for a shrinking pie. The companies that already rebuilt their visibility stack around what LLMs cite are pulling ahead with one-person teams. That gap will not narrow.

Writing in the Age of LLMs: Will the Explainer Survive?

The week's most interesting meta-thread was about whether the kind of writing these newsletters do still has a future. dynomight asked the question directly in "Will the explainer post go extinct?", working through the perverse incentive that any explanation you publish gets ingested into training data and then regurgitated, after which you can no longer claim to explain that topic better than the AI can. Copyright protects presentations, not ideas, and no one expects an LLM to credit its training data. The seven reasons he lists for why people still read human-written explanations are the interesting part, because they are the surface area on which human writers will have to compete.

The irony is that the inbox also contained a perfect example of the genre that is supposedly under threat: Aditya Bhargava at Ducktyped published "An Illustrated Introduction to Linear Algebra," a patient walk through Gaussian elimination using nickels, pennies, milk, and bread. It is exactly the kind of explainer that dynomight is worried about, and it is also exactly the kind of writing that the LLM versions never quite match in voice or sequencing.

signull ran the lyrical version of the same anxiety in "advice expires faster than milk now." The argument: the world updates faster than language can keep up, so received wisdom is now time-shifted before you finish hearing it. The recommendation is to "stop taking maps and start building a radar." It is the kind of piece that reads as either profound or empty depending on what week you read it, and this was a week where it read as profound. Ami Vora at The Hard Parts of Growth published the full catalogue of her writing, ordered for end-to-end reading, with a note that she will probably slow her posting cadence now that the initial list is out.

The take: the writers who depend on the explainer format have started to notice that their moat is thinner than it was eighteen months ago. The response is not to write less. It is to write the kind of post that the LLM cannot generate from the training data, which is the post that includes the writer's own headcount cut, their own attribution number, their own NotebookLM companion, their own catalogue. The defensible writing is the writing that documents a particular operator's choices in a particular week. The rest is at risk.

Macro Cracks: Subprime Auto and Red October

Two finance writers flagged the same underlying mood from different angles. The Last Bear Standing published "Here Comes the Repo Man," a post-mortem on Carvana's asset-backed securitizations. His claim is that loans packaged and sold over the past four years are realizing pronounced losses, that rating agencies are still clinging to stale assumptions, and that noteholders are either unaware or unwilling to force a public markdown. The losses, he argues, have already occurred. The market just has not priced them yet. Carvana itself is the second-order question.

Citrini published "Macro Memo: Red October," which is the bigger-picture version of the same instinct. His framework: the secular economy is roughly halfway through a capex cycle that shows no sign of slowing, while the cyclical and lower-income segments have been weak for some time. The Macroeconomic Ouroboros, in his terms, is the dynamic where wealth-effect-driven spending from high-income households masks underlying fragility in the rest of the economy. The K-shape is the structural feature, not the bug. He is also explicit that Fiscal Primacy means the administration will throw whatever it takes (tariff refund checks, MBS cap lifts on the GSEs) at keeping asset prices elevated through the midterms.

The take: the subprime auto story and the K-shape macro story are the same story read at different zoom levels. If Last Bear is right about Carvana, the consumer-credit cracks are already past the point where ratings catch up, and they will surface as a credit event before they surface in the official narrative. If Citrini is right about Fiscal Primacy, the policy response will be aggressive enough to delay the reckoning until after November 2026. Both can be true at once. That is the uncomfortable read.

Voice and Reflection: Birthday Notes and Thank-You Letters

The week's quieter pieces were also its best-written. Steven Schlafman at Where the Road Bends turned 46 and wrote about a breathwork session with Alice Wells in which he tried to describe how his awareness goes online during the day but vanishes in social situations, how his heart closes in group settings, and how the old armoring shows up around certain people, including his partner. The honesty is the point, not the framing.

Piera Luisa Gelardi at Noomalooma wrote about a Sassy magazine subscription that her mom's best friend Diane gave her at fourteen, which shaped her dreams of moving to New York, her music taste, and a decade later her approach at Refinery29 to making content that was playful and broke the rules. She finally sat down to write Diane a thank-you note in 2023. The piece is a small argument that the things you do for people land in ways you may never see.

Ben Kassoy announced he had started as a Senior Copywriter at FIGS. Annie Duke at Thinking in Bets released a podcast episode with Josh Wolfe of Lux Capital on the contrarian's guide to venture capital, organized around the question "what sucks?" as a way to spot opportunities others are not looking at. Guillermo Flor at VC Hiring published the week's 24 VC jobs and 60 remote-first companies. Upen at MicroSaaSIdea ran Issue #143 on $1K-$10K MRR products, with Manoj's Infography hitting $94K in 12 months and ReelFarm clearing $42.3K in July as the standout numbers.

Abby Falik at Taking Flight closed the week with "We're Building Fast, But Toward What?", a dispatch from the Masters of Scale Summit. The energy was palpable, the AI breakthroughs in gene editing and military tech and animal communication were everywhere, and the question that kept gnawing at her was Sandel's: how would an enlightened society approach this transition? She quotes Sean White of Inflection AI: "We'll get what we're aiming for. So, is the goal AGI? Or human flourishing?" Andrew Ng's answer at the same event, "Just go build, build, build," got thunderous applause.

The take: the personal essays this week made a better case for human writing than the explainer-post anxiety pieces did. Schlafman on his heart closing, Gelardi on Diane, and Falik on Ng's applause are all pieces that no LLM is going to ghostwrite well, because the substance is the specificity of the writer's own life. That is the moat, if there is one.

Three Takeaways from the Week

The operator economics shifted in public this week. Kohli published actual RouteMesh headcount numbers and Poyar published actual Docebo attribution numbers, and both pointed at the same thing: one or two people with the right stack are now generating output that used to require ten. The interesting question is no longer whether this is possible. It is what the cut five do next.

The writing economy is starting to feel the edge of the same blade it has been cheering. dynomight's explainer-extinction question, signull's expiration-of-advice lyric, and Ami Vora's catalogue-then-slow-down move are all early signals that the people who write for a living are reconsidering what they should be writing now. The honest answer, based on which pieces actually landed this week, is the writing that is irreducibly the writer's own.

If you only revisit three pieces from the week, I would suggest Kerman Kohli's "The Machine Economy" for the clearest founder-seat report on what AI is doing to team size, Kyle Poyar's Docebo breakdown at Growth Unhinged for the cleanest AEO playbook published so far, and Abby Falik's "We're Building Fast, But Toward What?" for the question that the rest of the inbox kept circling without quite asking out loud.