Using AI with and about Obsidian

⁠

Making Obsidian Work Like Me⁠

was written by

Claude.ai⁠

using the “Alexizer”, a Claude project that is configured with examples of my past writing. (Read more about the Alexerizer in my guide,

Inside You There Are Three AIs.⁠

) Alongside the prompt below, I gave Claude a Claude-generated summary of the day-long ChatGPT session where I conferred with GPT on my initial setup. So the AI-to-AI-to-AI workflow was:

Spent a day working in GPT to customize Obsidian

Exported my ChatGPT chat transcript as text using

Export ChatGPT⁠

(which I actually one of my core use cases for Obsidian: I’ve used it to export ChatGPT transcripts to Markdown files that I store in Obsidian. My long-term goal is to automate a weekly export from GPT to Obsidian so that I can work with my AI transcripts within Obsidian as knowledge files, which I should be able to query and manipulate with AI, too.)

Got GPT to write “gptwithoutcodeblocks.py” (see

Script Descriptions⁠

) to rip all the code blocks out of my GPT transcript so that it would be small enough for Claude to accept as an input.

Gave my consolidated file to a generic Claude chat with a prompt requesting an outline with detailed notes (see below). I didn’t go straight to the Alexizer because all the Alexerizer’s background files use up a lot of Claude’s room, so it won’t accept large files as inputs.

Took the detailed outline from generic Claude and gave it to the Alexerizer, with the request, “Write a draft in Alex's voice of "Making Obsidian Work Like Me". Use these notes as the basis.”

After reviewing the result (which you see verbatim in

Making Obsidian Work Like Me⁠

), gave the Alexerizer a further prompt, “OK, but that isn't detailed enough. Please include an appendix: The nerdy details. Use this for the very detailed recap of all plugins, scripts, etc. Assume it's a how-to guide for friends who are going to follow me down this path.” That’s what created

The Nerdy Details⁠

The prompt: Outline

Please take this log of a conversation of my efforts to get up and running with Obsidian and write a bullet-form list of points to include in my draft post, Making Obsidian Work for PKM. Be sure to note anything that is specifically an effort to address my 4 top pain ponts/goals:

Tool Overload and Fragmentation: I’m using a mix of Evernote, Coda, Scrivener, Zotero, ChatGPT/Claude, Skitch/XSnapper, and Obsidian/Tana. Each has its strengths, but the lack of integration and poor search functionality across tools is causing friction.

Search and Retrieval: The struggle to find information across these platforms is huge. Coda’s search is lacking, my Super Coda setup [using a Coda pack I wrote to create a directory of all my Coda docs] doesn’t quite do the job, and Evernote is becoming increasingly cluttered.

Mind Mapping and Visualization: I want the connective, mind-mapping power of tools like Obsidian and Roam, but their complexity makes them difficult to stick with. I’m drawn to Tana’s node-based system, but the lack of a mobile app is a major limitation.

Organization: I’m feeling overwhelmed by the disorganized state of your Evernote, the difficulty in managing everything, and the potential loss of serendipity that comes from having a well-organized system.

Make a list of:

Problems and frustrations encountered

Aha moments Strategies tried

Templates needed/created

Plugins tried, selected and/or customized: Which plugins, what are they used for, did I have any comments on them, what challenges did I face in set up that I asked about. For exmaple setting up a smart connect server.

Unresolved problems: What was I still struggling with by end of chat or did I set aside

Automations/workflows needed: what one-off fixes did I use that are going to require ongong setup/workflows . like a way of autostarting my local server for smart connect so I don't have to run npm start manually every time

A comprehensive list of all the python scripts crated with a note about what each one does. Return them in sequential order of creation. If there are scripts that got creatd multiple times, just create a sub bullet like "Had to create 4 versions of this script before I realized it was better if consolidated with [other script] to do both kinds of cleanup at once" or "Took 3 versions before it correctly tagged creation info"

Summary of level of tech skill needed.

possible costs like openAI API credits

Make sure to include that I should have backed up better before starting my experiments! sync isnt backup and obsidian is complicated that way

Want to print your doc?
This is not the way.

Try clicking the ⋯ next to your doc name or using a keyboard shortcut (

CtrlP

) instead.