Why I'm cutting our runway

Both of them are back from vacation, so this one starts with camping. Niklas did Norway in a small van with two kids, two car seats, two bikes and a stroller (lovely, but next year will be less ambitious). Peter camped too and spent about half of it getting kids to sleep in a tent. He also left his computer at home for the first time in years, which went fine apart from the odd thought about what would happen to it if something happened to the house. The best moment was sitting on a bench looking across the water at one of the most famous battlefields in Danish history and getting ChatGPT to explain the whole thing from exactly where he was sitting (that hill over there is where the long range cannons were, that island is where the Danes retreated to).

Peter's update is mostly memo. He's invited a bunch of people, most haven't really activated, and he's still using it a magnitude more than anyone else. He's made it automatic for himself with prompts and instructions, so Claude and ChatGPT just know everything lives in memo and they read, write and update memos on his behalf. The problem is that nobody else gets to the aha moment unless it's automatic for them too, where the agents fill up their memo in the background and one day it does something they didn't expect. He hasn't solved that yet, and he's not in a hurry, since he's learning a lot as user number one. The idea he's excited about is executable memos: a memo that runs at a certain point in the future. Start with reminders, and then notice that a reminder telling your Claude to do something in two weeks is just agent orchestration.

That ties into OG Kit, which Peter isn't really working on but people still pay for and like (Phone.inc became a customer this week). The thought experiment is whether an AI could run it, roughly what Sahil is doing at Gumroad but at much lower stakes. He'd start with the maintenance he doesn't do (updates, bugs, the backup that failed the other day), while Niklas argues he should start with marketing since that's where he does nothing at all. What Peter really wants is a layer above the pull request: something that fixes a thing, notices the fix broke something else, and either fixes it or rolls it back without any of it becoming his problem. Which leads to the open question he's sitting with: what building blocks would memo need before you could build the OG Kit operating system on top of it?

Niklas came back from holiday wanting to step harder on the pedal, cutting runway from eighteen months down to thirteen or fourteen and hiring faster (more product engineers, probably an engineering manager, and a marketing and growth lead before the public launch). He also made a roadmap, felt dirty about it, and then rebuilt the Phone.inc website with Cursor, ui.sh and Astro on Cloudflare Workers, feeding it 60 screenshots of sites he likes plus a monorepo with the app and the docs in it. Then he had an agent work for eight hours generating 25 integration landing pages (Shopify, QuickBooks and so on), and during a call with a potential integration partner he had one built, deployed and sent over as a link right after. He also did an AMA on the Danish startup subreddit that got 193 comments, 80,000 views and around 60 waitlist signups.

---

Still Early is sponsored by Phone.inc: a business phone number, welcome greeting, call routing, and opening hours, all from an app on your personal phone.