One Shot
Tings Newsletter #35
I’ve been agentic coding non-stop. Max plans on Claude and OpenAI with both subscriptions maxed out every week. 425 GitHub commits in the past month. I don’t see it slowing down. It’s the most fun I’ve had in a while!
I built Ideal Day for my wife using Claude Code. It lets you create travel itineraries, save favorite spots from your trips, and easily share recommendations with friends or visitors. I one-shotted it with a single prompt, then built features on top of it. I also ran her moodboard through Claude Design for the design. Next up: I’m working on a feature that will create guides for you based on your taste.
Some observations: intelligence is a commodity, coding is a commodity, distribution (sales & marketing) is a commodity. Even design is becoming one. The only real moat left is original thinking and taste. How do you create something that doesn’t look like AI slop? How do you generate original ideas that AI hasn’t already indexed?
Harnesses are moats. Intelligence is everywhere but the workflow, scaffolding, memory, and tools around the model are becoming a real advantage.
Building with these tools is both easy and hard. I probably spend half my time fixing things that break. Tip: If you have OpenClaw, Hermes, etc. working, don’t upgrade to the latest version. That way you don’t waste time troubleshooting and can focus on building instead. Without a background in building tech products, I would’ve quit ten times by now.
My current stack: Openclaw as the orchestration layer, Paperclip for task management, and OpenAI Codex GPT 5-5 and Claude Opus 4.7 for coding. I follow the thin harness and fat skills approach. And I keep my setup very boring.
Naval's take: a) pure software is becoming uninvestable, b) the programming language of the future is English, and c) the new moat is English + taste. Build niche, private, personalized tools for yourself — the kind nobody else would make for you. The big benefit of building your own software: you don’t have to compromise on your vision.
It’s a great time to be donkey-corn. Solo founders (or small teams) building $1M to $10M+ businesses. The unicorn path requires building SciFi tech and a $1T outcome. The donkey-corn path just requires taste and execution.
Tied to this: don’t get trapped raising a venture round if your idea is a donkey-corn. The outcome bar is way higher in a post-AI world and building pure software is unlikely to clear it. That said, if you do end up building software, make sure it has data network effects.
Every overpriced SaaS will eventually get a free open-source equivalent. It’s just a matter of time. For example, on Docusign: a “10-person team on Business Pro pays $4,800 to $7,800 a year. To put signatures on PDFs.” DocuSeal (open-source alternative) does the same thing for... free.
Open-source models are catching up to closed-source models. Frontier will always be frontier, but the leaps between the two keep shrinking. See the 2026 AI Index Report.
Is Bitcoin susceptible to quantum? Google just set 2029 as their internal deadline to migrate to post-quantum cryptography (PQC).
Chinamaxxing is Gen Z’s word of the week. Drinking hot water, wearing house slippers, and listening to Joji. The deeper read: it’s not really about China. It’s about how broken America feels right now.
And the Freakonomics version of the same story: China is run by engineers, America by lawyers. We’ve optimized for litigation. They’ve optimized for shipping.
Updated my simulation theory. I used to picture it as living in a video game. New theory: a future civilization is going extinct, and they’re running infinite simulations to find the right path forward. Statistically, we’re in one of those. Not base reality.
Lp(a) update: the 2026 ACC/AHA guidelines now recommend Lp(a) testing for every adult, once. If you’ve never been tested, get tested.
People are no longer using GLP-1s just for diabetes or weight loss. They’re using them for other conditions like arthritis, addiction, long COVID, IBS, anxiety and other inflammation-related issues. The bigger story may be that GLP-1s don’t just affect appetite; they appear to act on the brain, reward pathways, metabolism and immune signaling.
Tech guy in Australia adopts a rescue dog with months to live. Pays $3K to sequence her tumor, feeds it to ChatGPT and AlphaFold, designs a custom mRNA vaccine. Tumor halves. Dog is alive. “If we can do this for a dog, why aren’t we rolling this out to humans?” We are going to cure so many diseases this decade.
Caught the knitting bug. Just finished my first scarf. Apparently the whole Gen Z boomer-hobbies thing is a real trend with knitting, gardening, and baking. I’m also working on getting this cookie recipe.
board.fun for family game night. Adding a few of these to the rotation alongside Sea, Salt and Paper.