welcome to my world :)

I wanted to understand more of the world without collecting an endless list of things to read. Pillars curates papers, articles, and essays each week, shaped around my interests and the gaps in my knowledge.

There are five readings and a Sunday revision. The week moves between foundational ideas, current thinking, connections across subjects, and something chosen for the pleasure of it.

The learning loop borrows from the Feynman technique: try to explain an idea in plain language, and notice where the explanation falls apart.

I can save an explanation or note what still feels fuzzy. Those loose ends shape later revision questions and future reading. The knowledge map grows as I read and recall, showing the subjects I return to and the areas I’ve barely touched.

It’s an iOS app I built for my own learning, with a map that keeps changing as I do.

I fed a living wiki my exported chat history, emails, and documents. It extracts concepts and links them together. The chat history gave it the range the other sources were missing.

Everything goes into three buckets:

Core

What I care about, return to, and am known for.

Spine

The values and positions that hold it all together.

Shadow

What I reject, avoid, or refuse to engage with.

The shadow turned out to matter most. Knowing what I’m not made its predictions of what I am much more reliable.

I asked whether I’d laugh at a misogynist joke in a room that was 80% men. It read across my humour, values, and notes on how I act in a room, including the less flattering parts.

“I’d just go flat. Blank face. Not laughing reads loud in a room of 80% men, that’s the whole point.”

That’s exactly what I’d do. Seeing it find that answer was the first time I thought this could work.

This started with a question about social networks: could a version of you find the people and information you care about, without you spending your day on a feed?

I called the matching concept the Social Dark Pool: representatives test a conversation privately before bringing an introduction back to the people involved.

The clone is one piece of that idea. It’s a model of one person; it doesn’t prove that two people would connect. I needed a smaller matching problem I could actually test, so I started with my taste in clothes.

“Summer in Berlin” should mean different things to different people. Styled starts with a taste graph: silhouettes, colours, brands, references, and the price I’ll actually pay. It includes the things I’d never wear.

Each new item gets a compact description in the same format, so the system can compare the item with my taste before asking a model to reason about it.

Rules remove the obvious misses: sold out, wrong category, a colour I won’t touch. Image and text embeddings narrow the rest by visual fit, brand, season, and novelty. A diversity pass keeps the shortlist from becoming twenty versions of the same jacket.

Only then does a vision model look at the remaining images, my taste graph, and the last twenty things I saved. It picks the final set and explains why. The expensive reasoning comes at the end.

I can tell within seconds whether I’d wear something. That gave me a way to test the filtering architecture on thousands of inputs, with a result I could judge myself.

It tests personal taste matching. Whether two people would get along is a different question, and still an open one.

I built a system that studies creators’ videos, the words on screen, the opening, the cuts, and the comments underneath. It maps what happens moment by moment, so I can ask why an opening works and get back the source.

It also compares a post with that creator’s usual performance. A big view count alone doesn’t tell me much.

The useful part is connecting an audience question to something I can make: a small tool, a tutorial, a story, or a new way to explain an idea. Each suggestion needs a first frame, something worth watching for, and an actual result for the viewer.

This is an ongoing experiment inside Lola OS. I’m still testing which recommendations deserve to become posts.

I send links, notes, and voice messages to one Telegram chat. The bot reads or transcribes them, saves the original, and links the ideas to things I’ve saved before.

I can ask it a question later without remembering where the thought came from. It searches a persistent knowledge graph, with a readable version in Notion.

It has a memory across conversations. An article about physics can connect to a design reference, something I’m writing, or a conversation I had months ago.

The same system now supports my creator research and a personal newsletter that draws from my interests. It’s a private tool I use and keep developing.