Hyper-learning with Claude
Dec 11, 2024
I’m experimenting with learning a very broad and dense space fast: why home ownership is so out of reach for people in the United States. I want to learn a couple of things:
- What are the drivers of “not enough housing”? Targeting 5-10
- Hypotheses on which are most tractable
- A list of companies (or build my own?) that are trying to target the most tractable
In order to do this, I’m leaning on Claude, ChatGPT, and Perplexity to help me. In this post I’ll list the workflows that I am finding a lot of alpha with
1. Doing the reading
This is the single biggest lift: I will download 4 dense housing research papers to pdf, and then upload them into a Claude conversation. Claude can then crunch all 4x30 = 120 pages and help me quickly understand what part of the giant housing mental model landscape these papers are talking about
I will typically group papers together—usually by the think tank or non-profit that I find them on. I hope to clash papers from different organizations together, but right now I haven’t filled in enough of the housing mental map to say “oh these organizations hold unaligned positions about [topic]”
I’m also experimenting with an Arc Boost/Chrome Extension to one-click dump to PDF to make this even faster
2. Folding more concepts into my mental model
One of the trickiest parts of learning a new space is what I call “building the tackle box”
I’m experimenting with giving Claude my tacklebox, and asking it where it would slot the concepts that come out of each of the reading sessions I describe in (1) above
I haven’t cracked this one yet, but I am hopeful
3. Maintaining and feeding Claude a “Dump Doc” from time to time
This is analogous to the tacklebox, but just more verbose. I have a Notion doc that has pages and pages of raw text and images that I thought might come in handy. Occasionally I’ll feed it into Claude, and ask it to suggest things in this conversation that I should consider adding to the Dump Doc
The goal here being to maintain some semi-refined, but still raw, form of housing data, hypothesis, research, and well-informed opinion that I can keep returning to