# Dive52

A weekly learning loop that turns articles, podcast clips, and voice reactions into something I can remember, question, and use.

[Read the story ↓](/content/projects/dive52#why-i-built-it/index.html)

## Why I Built It

I can read 30 articles in a week, flag the interesting ones, and by Sunday remember almost none of them.

The missing step was not more capture. It was a forcing function between consuming an idea and deciding what I think about it.

## The Loop

Each week gets one topic. Articles, podcast clips, and voice notes land in Discord, where I react to the ideas worth keeping. Claude uses those reactions—not just the raw reading list—to create a structured synthesis in my Obsidian vault.

01

Choose

One question for the week

02

Collect

Articles, clips, voice notes

03

React

What matters, surprises, or connects

04

Synthesize

A durable point of view

The constraint is the product: not “learn more about AI,” but “this week, what is actually happening with AI data-center buildout?”

## What Makes It Useful

- Discord keeps the input and reactions lightweight
- Heart, star, bookmark, and skip reactions tell the system what I actually care about
- `/ask` lets me question the week’s source material while it is still fresh
- `/synthesize` produces the essay; `/publish` saves it to the vault

## Status

The system has fetched 512 articles and generated 215+ structured cards. It runs continuously for about $3 a month. The useful output is not the volume—it is having a point of view I can find again.

## Tools that helped

Claude  Obsidian  TypeScript  Multi-agent system

## Up next

Permit Propensity

I’m investigating whether public permit and real-estate data can uncover new opportunities and better leads for high-end home service businesses.

→](/content/projects/permit-propensity/index.html)