Week 46 · 2025-11-10 → 2025-11-16 · 22 newsletters

Vibes, Benchmarks, and Kindness

ai-for-builders · saas-state-of-play · craft-and-kindness · storytelling-and-power

A quieter week in the inbox: 25 emails after filtering, mostly Substacks, with no single dominant news event. The throughlines were AI tooling at the practitioner layer, the state of SaaS at the end of a long year, a small cluster of writers thinking about craft and kindness, and one outlier on Trump Organization deal flow. Nothing breaking, but a few pieces worth saving.

AI for Builders: From Vibe Coding to Vibe Designing

The week's most cohesive thread was practitioners trying to figure out what the AI tooling stack actually looks like now that the hype has hardened into workflows. Marily Nika at her AI Product Academy ran a long hands-on guide to Google AI Studio, framing the build cycle as having shifted from "Idea, PRD, Debate, Refine, Build" to "Idea, Brainstorm with AI, Prototype the vibe, Team experiences it, Refine, Build." Her case for AI Studio over v0, Bolt, or Lovable is that the Gemini stack ships native, zero-configuration access to Gemini 2.5 Pro, Nano Banana for image generation, and Veo 3.1, all in one place.

The companion piece came from Xinran Ma at Design with AI on Vibe Designing, the argument that Karpathy's "vibe coding" framing breaks down for product designers because the vibe coding workflow is linear and design exploration is not. Her observation is small but worth keeping: Google Trends shows "vibe coding" exploding while "vibe designing" barely registers, which is a real gap if you believe the design exploration phase has any value at all. Carilu Dietrich wrote up seven drivers behind Lovable's growth, and the most honest line was the meta one: you cannot copy Lovable's tactics because their momentum is itself the multiplier. The lessons that actually generalize are "ride a macro wave bigger than you" and "best-in-class social plus deep community."

Ethan Mollick at One Useful Thing ran Giving your AI a Job Interview, which is the right counterweight to the tooling pieces. His core point: most public benchmarks are flawed (MMLU-Pro asks about Cheap Trick albums), but taken together they still measure some underlying ability factor that trends up and to the right. The practical implication for builders is that benchmark scores are a noisy signal you should not over-index on, but the trend line across many noisy signals is the thing to trust.

George Mack at High Agency ran the most readable piece of the week, How to be creative (without taking drugs), a list of twelve inputs that increase creativity. The "translate aggressively" rule (draw a written idea, write a visual one, explain an equation out loud) and "scroll for anti-social proof" (only click YouTube and Substack content with under 5,000 views) are the kind of cheap interventions that compound.

The take: the AI tooling conversation has moved past "which model is best" and into "what does the workflow actually look like for my job." Marily and Xinran are running the same play from opposite sides, and the gap they expose between coding and designing is the one tool builders should be racing to close.

SaaS State of Play: Benchmarks, Pricing, and a Long Tail

Kyle Poyar at Growth Unhinged ran the 2025 SaaS Benchmarks report in partnership with High Alpha, with over 800 B2B SaaS companies participating (a new record). The headline framings he flagged: the Efficient Growth matrix is the metric to watch, not LTV:CAC; AI-native startups are not behaving like B2B SaaS in their growth curves; engineering is the function getting hit first by AI on the headcount side; and there is a real pricing sweet spot showing up in the deal-size data. A few days later he followed with How to use AI agents for marketing, a case study on SafetyCulture's GTM engineering team showing near-100% lead enrichment coverage, 2x increase in opportunities created, 3x increase in meeting booking rates from AI-powered outbound, and 10% lift in feature adoption. The line worth flagging: "over 90% of marketing teams use ChatGPT as their main AI sidekick. Relatively few have agentic AI workflows in production with customers." The gap between "we use AI" and "we have agents in production" is still enormous.

Upen at Micro SaaS Idea ran the part-two roundup of $1K to $10K MRR products, and the names that stuck were Vocal.email ($12K/month for voice notes in Gmail and Outlook), Oh Dear ($1M ARR for website monitoring), and Blitzit ($26K/month). The pattern across the long tail is unchanged: small, specific, boring, profitable.

The take: Poyar's benchmarks plus the SafetyCulture case study are the cleanest read of where B2B SaaS actually is right now. Agentic workflows in production are still rare enough to be a real edge, and the operators who are running them are seeing real numbers, not vanity ones. If you are a GTM leader, the SafetyCulture playbook is the one to study.

Craft, Kindness, and Delight at Work

A small cluster of writers landed on the soft skills end of the spectrum this week, and the pieces talked to each other better than I expected. Teiva Harsanyi at The Coder Cafe wrote Nothing Beats Kindness for World Kindness Day on November 13, with the on-call colleague story as the anchor. The argument is operational, not sentimental: kindness creates psychological safety, psychological safety surfaces risks earlier, and risks surfaced earlier ship faster. Nesrine Changuel at Product Delight Tips ran the Product Delight Model applied to travel, with the three-tier framing of Low Delight, Surface Delight, and Deep Delight as the model. Travel is the right domain to test the framework on because the emotional stakes are unambiguous.

Method Studio put out a video compilation on creative direction habits from Kevin Mulroy at Mischief, Hal Curtis and Felipe Ribeiro from Wieden+Kennedy Portland, and Guillermo Vega at Ogilvy. Wendy MacNaughton at DrawTogether ran a Native American Heritage Month piece on three contemporary Indigenous artists starting with Wendy Red Star. And Richard King at The Product Marketing Drop wrote Why your product release emails get ignored, with the practical reframe being that subject lines should lead with outcomes, not features ("Record videos without being on camera" beats "Introducing Our New AI Avatar Feature").

The take: kindness, delight, and craft are not soft topics, they are the operational substrate that makes everything else work. Harsanyi's piece on kindness as a speed multiplier is the one to send to engineering managers.

Storytelling, Democracy, and the Trump Organization

Peter Teague at The All American ran two essays from the Out of Many, One anthology: Mari Manoogian on universal opportunity as the renewable American promise, and Steven Olikara on storytellers as the front line of democratic repair. Olikara's framing, that politics lives downstream from culture and the WWE-style Red Team Blue Team narrative is "addictive for many, profitable for some, and disastrous for our democracy," is the right diagnosis.

Zach Everson at 1100 Pennsylvania returned from a hiatus with a long Trump Organization update: Eric Trump telling a UAE crypto conference that the Trump Organization is "doing projects in Jeddah, Oman, Riyadh, all over the region," which is a notable distance from the 2017 pledge of "no new foreign deals whatsoever." Abu Dhabi's MGX fund is now a key player in two major Trump-era deals, including a $2 billion Binance deal denominated in the Trump-linked USD1 stablecoin. The piece is worth the click for the receipts alone.

The take: Olikara and Everson are doing different versions of the same job, naming the gap between the stories Americans tell themselves and the deals being cut in the open. The fact that both pieces landed in the same week is the kind of coincidence that says something about the news cycle.

Three Takeaways from the Week

The AI tooling conversation has shifted from "what model" to "what workflow," and the Marily Nika and Xinran Ma pieces are the clearest read on where the practitioner gap actually sits. If you are building product right now, the vibe-coding-versus-vibe-designing distinction is worth taking seriously.

The B2B SaaS world quietly hit a real inflection point on agentic AI in production. Poyar's benchmarks plus the SafetyCulture case study suggest that the operators running agents are seeing 2x and 3x lifts on real funnel metrics, and the 90% of teams still using ChatGPT as a sidekick are about to feel the gap.

If you only revisit three pieces from the week, I would suggest George Mack on creativity for the cleanest practical frame, Ethan Mollick on benchmarking AI for the cleanest read on where measurement stands, and Teiva Harsanyi on kindness for the post that will quietly improve how you show up to work next week.