In AI, approaches live short lives — curious which ones survive.
One that works for me now: a component catalog with usage counters, a changelog agents write themselves, and guidelines they read first. Core Platform, scaled down to one product.
https://t.co/ZiTdqcxVOw
I’ve updated the diagrams in my article about the Core Platform with AI. The ideas are still relevant, and this week I’m planning another short post about the Operations Layer and how it applies to the platform I’m actively working on.
https://t.co/O59dOo0riM
A short post about how I generated a collection of bugs. No tutorials, just lots of images and a solid result. I’m impressed with the outcome and how quickly I was able to create a whole collection in a consistent style.
https://t.co/FJyI2UmMTn
Wrote a short post on my blog about automating Core Platform with AI agents.
The routine that used to require a dedicated team - validating tokens, tracking contributions, publishing platform health - now runs with three agents in the background.
https://t.co/9SKFT1LndE
@ericmigi Using Apple Watch Ultra and Garmin Fenix now, had various Samsung watches before. No new scenarios that Pebble would close better than these. Still want one badly though — purely for the vibe and the ability to write my own software 🤟
In my case, I built a personal finance system that pulls data from several banks I use. On a regular basis it sends reports into my Obsidian Wiki, analyzes progress toward financial goals, generates recommendations, and helps with weekly planning.
My recommendation: when building these kinds of tools for yourself or your team, integrate them with a Second Brain. It adds an intelligent analytical layer that helps you organize your work and optimize projects at a deeper level.
@jameesy Excellent point - using two vaults is a great solution. I ran into the same issue: with the LLM-maintained wiki approach, my single vault got heavily overloaded, and navigating my own notes became much slower and less intuitive than before.
@stitchbygoogle That's awesome! I especially love the bottom-up approach. Starting with the most basic elements, then moving up to components, blocks, and finally concepts - it feels really logical and scalable.
I kept seeing this “Wiki LLM + Claude + Obsidian” setup by Andrej Karpathy. https://t.co/XaiiDvc7LU
At first — skeptical. Then I tried it.
It blew my mind 👇