If you've read a few of these, you know I have two newslettering modes: a short note and a bunch of links, or a long essay. I had planned to do Mode 1, but ended up writing a Mode 2.
But, as I care deeply about you, reader, and the user experience of this newsletter, I'm gonna put some links up top, and you can keep reading on to find out about a VC event I went to and how MBAs have appropriated the language of engineering to sound smart.
The caviar, it turned out, was a psy-op, and everyone's "future-proof" skills were "sales".
π The Curation Problem β Related to the above, Ibrahim Bashir's latest newsletter is about how, now that AI has accelerated delivering products, we risk losing the plot on product discovery, confusing activity with impact, etc.
The takeaway shouldnβt be that you can ship more than ever - at some point, your userβs ability to even absorb changes gets saturated. And the dangerous mindset I see many organizations adopting is ship all the things: every random idea, every hackweek prototype, every edge case, every bell and whistle. β ββIbrahim Bashir, The Curation Problemβ
π€ Scaling delivery with AI agents β In this post, I describe my version of an agentic software factory. (There are many like it, but this one is mine.) More specifically, it's about how I'm using AI tactically to get more done, while reserving plenty of spots to test and review the output and ensure my clients are getting excellent work.
ποΈ Using Sanity for internal tools β I can't believe I spent most of the last one on Sanity and didn't mention this post. π I've been building lots of small internal tools for bits of my client experience that weren't quite right using off-the-shelf apps. And, since a lot of these use cases involve text or structured data, Sanity has been a natural fit for the backend.
Scene from the Primary NYC Tech Summit Founders Dinner at Pier Sixty, September 15, 2026.
Everyone who should be here is here
I spent Tuesday night and half of Wednesday this week attending Primary's NYC Tech Summit, billed as "1,500 of the most consequential builders and backers in one room, for one day." This was a free event β or at least I somehow scored a free ticket β and seemed more interesting than being alone in my office, so I went, thinking, well, at least there will be builders? Fellow techies? Maybe?
I'm sad but not surprised to report that the tech content at this "tech summit" was nonexistent. This was a business event for money people, and what's more, it drove home something I've been noticing for the past few years but hadn't been able to describe until now:
Now that the world is fully saturated with computer and internet tech, and as such tech progress has slowed and maybe even stagnated, the business side of tech β marketers, MBAs, "operators", etc. β has fully taken over our industry.
What's more, tech business people have appropriated the language of engineering in much the same way they already toss finance jargon into sentences to make them sound more technical and profound.
You might think "builders" are people making tech products, right? Nope. They're the people funding tech products, either hustling to raise capital so they can hire people β or AI agents β to make the actual product, or (worse, yuck) the VCs allocating said capital and then bloviating on Twitter.
It's become common for tech-business people to use words like "stack" or "operating system" to refer to the tools and processes by which they run businesses. Sometimes an "operating system" is a product, like a combo CRM-project management app for dentists or hair salons. Usually, it's just a vibe, e.g. a sales consultant talking about how they've built a "scalable OS for cold outbound."
"Scale" is, of course, another borrowed word. In a software engineering sense, to scale is to solve the thorny problem of making data or services more available and reliable, so you can serve more customers without making things slow or showing a Fail Whale. In the business context, to "scale" is just to make a company bigger.
In a related trend, a couple of weeks ago Georgia Witchel wrote about tech β specifically, San Francisco's startup scene β having a hot girl renaissance.
What's been happening: some startups, most notoriously Corgi (which ostensibly sells business insurance), have been hiring lots of young women to work on their growth/GTM teams, where they do a lot of networking, throw a lot of events, and post a lot.
As Georgia points out, the "Corgi Girls" model seems on the surface a lot like the old practice where tech companies hired "booth bunnies" to attract folks at trade shows. But the women working in startup GTM are actually selling their brands/products, not just using their youth and looks for attention β in other words, they're doing sales.
Hiring fleets of hot women as salespeople has been the norm in certain industries (like pharma) for decades. What this SF mini-trend indicates is that, in a post-scaling era for most of tech, even our scrappiest new startups' marketing motions have more in common with Big Pharma than early Google or Facebook.
For me, this pairs almost too well with tech-business's appropriation of software language to make something really boring but powerful (capital allocation) sound like tech companies are still innovating and changing the world, in that it's all vibes. You can have a "tech" company that primarily sells insurance policies, where on one hand the founder is talking about scaling and building and systems, and on the other, their sales team are throwing parties for founders and racking up impressions on X, all of which supports and grows the brand.
Tech in a post-tech era
When it becomes the norm for people to just start using technical language divorced from its original meaning, it shows how little tech business leaders understand and respect the technologists whose words they've appropriated.
And this becomes very salient with AI β if a "builder" in the business sense believes they can now also build products in the engineering sense, why have engineers? Why talk to users? Why not just have Fable or Astra go build an operating system that consumes tokens and prints money?
Quick aside: Some people on Twitter have literally tried this experiment, giving agents corporate credit cards, phone numbers, and access to ad networks, with the goal of having AI autonomously create businesses and make money. In one case I saw recently, an agent made 10,000 cold calls and made a staggering $75. (No idea how much the AI compute/inference cost, but it was probably more than $75.)
Back to the NYC Summit, in a twist I'm tempted to call ironic if we weren't living in a post-irony era, it turns out that 1,500 "builders and backers" can't fit in one room.
Adding to the post-truthiness of the function, every seat at the Tuesday night founders dinner had a tin of Russ & Daughters Caviar. But it was a psy-op β it was chocolate caviar, which is to say, the little chocolate cookie balls some places put on ice cream. It would have been such a nice touch if dessert had been ice cream, and we were instructed to garnish it ourselves with the 'caviar', but alas, dessert was tiny cakes and the tins were just party favors.
None of this was disastrous enough for the organizers to have to acknowledge issues or treat the day as anything other than a raging success. Honestly, for most of the builders and backers, it was probably great, though I bet like me they'd have preferred sandwiches.
The big question I'm left pondering β aside from whether I should have started a company building an AI operating system for deep sea fishmongers, or some such β isn't "what's next?" so much as "what is real?"
Though, actually, they're the same question. The best predictions of the future are often the smartest extrapolations from the present, ones that look past the trends we want to continue, and consider the trends that are most likely to sneak up on us.
Figuring out what today's ground truth is beneath all the AI noise and post-scaling tech hype is the real work for all of us who've made a life helping improve people's lives through technology.
With that, I leave you with another picture of Johnny:
