Week 39 · 2025-09-22 → 2025-09-28 · 15 newsletters

Risk, Verticals, And Witnessing

ai-as-decision-proxy · vertical-ai-and-consumer-to-enterprise · founder-risk-and-career-choice · going-to-see-the-physical-thing

Fifteen emails across seven days. A sparse week, the kind that rewards reading rather than skimming. No single news story drove the inbox; instead a handful of independent essays converged on a few real through-lines: AI creeping into decisions that used to be ours, the consumer-to-enterprise AI playbook starting to look like a real pattern, founders and operators reframing risk and career choice, and one piece that argued the only honest way to understand the AI buildout is to physically go look at it.

AI as Decision Proxy: When the Outsourcing Stops Being Cute

Two pieces hit this thread from opposite directions. Alex Furmansky at Magnetic Growth opened the week with a confession piece about letting ChatGPT talk him out of flying across the country for a third date. The model's advice ("match investment to reciprocity, avoid pressure, stay in the confident-grounded lane") was, in hindsight, correct. The unsettling part of his essay was the next beat: if AI was right about the date, why wouldn't you let it write your strategic recommendation, your difficult email, your boss-facing analysis? The boundary between "tool that helps me think" and "proxy that thinks for me" turns out to be a slope, not a line.

Carilu Dietrich at Hypergrowth Leadership ran the marketing-side version of the same anxiety. SEO is tanking, she argues, and Answer Engine Optimization is the replacement discipline. Her interview with Profound's Nick Lafferty surfaced the operational shift: prompts replace keywords, Wikipedia and Reddit now rank in LLM responses far more than they ever did in Google, and new pages can get indexed by an LLM in 48 hours. The marketing function is being rebuilt around the fact that the customer's first read of you is now mediated by a model, not a SERP.

The take: the AI-decision-proxy story and the AEO story are the same story at different altitudes. The interesting question is no longer "can the model do this," it is "what part of my judgment am I quietly handing over, and is that the part I should be handing over?" Furmansky's date worked out. The strategic recommendation passed off as your own analysis probably will too. The bill comes due later, in the muscles you stopped using.

Vertical AI and the Consumer-to-Enterprise Bridge

The week's strongest industry cluster came from two VCs writing about the same pattern from different angles. Kyle Poyar at Growth Unhinged framed it as "the rising consumer to enterprise AI playbook," anchored on Fyxer, the Cursor-for-email startup that went from $1M to $17M ARR in eight months and just closed a $30M Series B led by Madrona. His larger argument: the old consumer software model was built for churn and tiny ARPU. AI is breaking both, with breakout consumer AI companies reaching $100M ARR in under two years and ARPU potential an order of magnitude higher. The playbook he sees forming: consumer speed for top-of-funnel adoption, then a bridge to enterprise for the stickiness and the contract value. The PLG era, re-engineered for the era of viral demos and ChatGPT-as-therapist.

Nikhil Basu Trivedi at next big thing ran the vertical AI version. His Waldo post was nominally a Series A announcement (Footwork led, Boldstart and Alt Capital and Sunflower and The New Normal Fund participated), but the framing is the broader one: the race is on to be the AI leader in every vertical, and the winners will be the teams that combine technical depth with deep domain expertise. Harvey in legal, Abridge in healthcare, OpenEvidence in medicine, Confido in CPG, Elicit in scientific research, WindBorne in weather, Waldo for agencies and brands. The thesis is that enterprise buyers are more primed than they have ever been to rip up their tech stack in favor of AI-native, vertical-specific tools, and the founder who knows the workflow beats the founder who only knows the model.

The take: Poyar and Basu Trivedi are describing two sides of the same arbitrage. Consumer AI is the new demand-generation engine for what eventually becomes vertical enterprise software. The interesting startups in this period are the ones that can ride a consumer-style top-of-funnel into a vertical-specific enterprise wedge before a horizontal incumbent notices.

Founder Risk and the Career-Choice Frame

Three writers, three different angles on the same instinct: stop optimizing for the wrong thing. signull at Signull vs Noise wrote the most provocative version. Every startup is a bet on mispriced risk, and the founder's job is to identify the risks they personally can take that others can't. Low burn (no kids, no debt, friend's couch for 18 months), cultural fluency (grew up online, speaks API natively), taste risk (obsessed with an idea 99% of people find cringe). The line worth sitting with: "you are not a founder. you are a portfolio manager of asymmetric, illiquid, non consensus" bets. Big companies are optimized for risk reduction by structural necessity, which is precisely why your edge as a founder is the risks they cannot take.

Ami Vora at The Hard Parts of Growth ran the operator-side companion. Big job decisions feel like they will determine your whole career, so you build elaborate spreadsheets in your head and listen to people who tell you to optimize for your "long-term goals." Her honest read on her own career: the roles that turned out to be inflection points were almost never the ones she would have predicted. The frame she lands on is "optimize for feeling lucky," meaning take the role where you suspect good things might happen in ways you cannot fully name yet. It is the calmer cousin of the signull argument: you are still placing bets, just on a slower clock.

Brianna Zuniga at Circular Architect, recently landed in her "dream role at her dream firm" (776's Angel Squad), wrote the most personal version. Three days at an investor event in Chicago, and what she actually noticed was that the venture world, supposedly optimized for capital and scale, is full of people clawing after meaning. The piece is sentimental in a way she earned, and the closing move (circling back to the "ecstatic, weird little girl who dreamed big" after a year of trying to outrun herself) lands because it admits the loop that the signull and Vora pieces only describe from the outside.

Paul Stansik at Hello Operator ran the small, useful counterweight. There is no secret to growing faster. There is no hack. You pick a niche, you understand the people who work there, and you blanket them with the most helpful stuff they have ever seen. The work, he argues, is to "hang around the hoop" so you are the first call when the buyer is ready. The piece reads as a deliberate antidote to the founder-as-risk-arbitrageur framing: most of the value comes from boring, consistent, useful work.

The take: this is a thread the inbox keeps returning to in different costumes. The signull and Vora pieces are aimed at the same person at different life stages. Zuniga is the after-photo of someone who took the leap. Stansik is the reminder that even after you have taken it, the daily work does not get more exotic. The honest read across all four: your edge is structural (what you can risk that others can't), your forecasting is bad (the lucky roles were never the ones you planned for), and the work itself is unromantic.

Going to See the Physical Thing

The single piece that did not fit a cluster was also the week's most memorable. Citrini Research sent a drone to Abilene, Texas, to look at Stargate. The framing was an argument against pure screen-based investing in the AI data center buildout: "chips don't float, fiber doesn't spool itself, power doesn't magically appear at the edge of the rack." So they partnered with Hunterbrook Media (the same investigative outfit they used for the Teradyne thesis) and went to physically witness the megawatts meeting the mud. The piece is a quiet reminder that the AI infrastructure trade is, at the end of the day, a real-estate-plus-power-plus-cement trade, and the people who win it are the ones who have actually stood next to it.

The take: the dynomight shorts essay this week (Shoes, Algernon, Pangea, and Sea Peoples) made the inverse point in a different register, opening with a riff about how we are in the "waning days of the People Read Blog Posts About Random Well-Understood Topics Instead of Asking Their Automatons Era." Both pieces are arguing the same thing from opposite ends: as more of the world becomes mediated by models, the value of direct contact with the actual thing (the field, the blog post, the megawatts) goes up, not down.

Three Takeaways from the Week

The decision-proxy slope is the AI story underneath all the other AI stories. Furmansky's date and Dietrich's AEO piece are about the same shift in different domains: the model is now the first reader of your situation, and the boundary between "tool" and "proxy" is softer than anyone wants to admit. Worth watching how your own usage patterns drift over the next quarter.

Consumer-to-enterprise AI is becoming the dominant pattern for breakout startup outcomes, and vertical depth is the moat. Poyar's Fyxer case study and Basu Trivedi's Waldo announcement are two readings of the same playbook. If you are building or investing in this period, the question is whether your team has a credible bridge from consumer-style adoption to enterprise-grade contracts, and whether your domain expertise is deep enough to outlast a horizontal incumbent waking up.

If you only revisit three pieces from the week, I would suggest signull's "building a startup is all about underwriting mispriced risk" for the cleanest founder-risk frame of the year, Kyle Poyar's "rising consumer to enterprise AI playbook" for the cleanest read on where breakout AI startups are scaling from, and Citrini's Stargate field trip for the reminder that the AI trade is a physical thing in a field in Texas. Three pieces, three reasons to slow down.