Enabling Environment
Imagine the place you do your best work. Andy Matuschak defined this, and as someone who aspires to effectiveness, I fell in love with the concept. As a techie who loves efficiency, I also set about building my own. From Andy: An enabling environment significantly expands its participants’ capacity to do things they find meaningful and important. This blog is the consumption environment for mine. When we talk about an AI enabling environment the key is not to figure out how you can prompt the AI, but how the AI can prompt you.
- From Andy
- Igor’s Computing Setup
- Igor’s Physical Setup
- By Task
- The rider, the elephant, and the path
- By Scope
- Visualizations
- AI Augmentation
There’s so much to say here. Pulling in some content from Andy:
From Andy
Igor’s Computing Setup
- Igor’s about page
- Some of the features in the readme.
- Tons of VIM/CLI stuff
- My general vim notes
- Vim specifically for writing
At its core, is writing:
Igor’s Physical Setup
- I think there are some coffee shop rituals that are part of mine.
By Task
Production
Consumption
- Having custom made podcasts can be better then books.
Reflection
Discovery
Recall
The rider, the elephant, and the path
You’d think this is about the rider, but the elephant is super important too
Inspire and Motivate
Mood matters, and for some reason, illustrating my blog makes me thrilled. If I’m stuck having DALL-E illustrate my blog makes me thrilled!
Check out illustrate.py
By Scope
Individual Artifacts
Check out think.py
Entire corpus
Check out this thing on embeddings. I think the story here needs to be embeddings + rag
Comparing Artifcats or Corpus
Visualizations
Word clouds
Old school, but pretty simple, perhaps do a second pass through gpt to simplify
Comic illustrations
Dimension Reduced Vector Embeddings
Connected Graphs
Ala - obsidians graph view
AI Augmentation
What I’m most excited about is how to expand this with AI. My thinking program is my first cut.
Spell check and summary/eli5/TLDR are the obvious ones, but what if we extended to reflection questions, implicit assumptions, etc….
The Danger of AI
The danger is of course you stop thinking. Learning requires active engagement with content, passive consumption, or even worse skimming makes things much worse. So it’s very important that you think about how to have the AI prompt you to be active in your thinking, not passive in your consumption.
Completing tasks vs getting smarter/stronger
There’s an analogy to physical strength. You can have a robot lift something for you, which may not be what you’re looking for. What you’re often looking for is how to get stronger.
To get stronger, you need progressive overload, which has different models from physical health:
- Increasing weight or resistance
- Increasing repetitions
- Increasing frequency of training
- Decreasing rest periods between sets
- Changing exercise variations
Similarly, for cognitive tasks, you can progressively overload by:
- Increasing complexity of problems
- Tackling more abstract concepts
- Expanding the scope of your knowledge
- Reducing reliance on external aids
- Applying knowledge in novel contexts
The goal is to challenge yourself just beyond your current capabilities, promoting growth and adaptation.
Critical Thinking
Running Think over Andy’s enabling environment page …
Historical note - original run on the model took like 60s and was dumber
github.com/idvorkin/nlp/blob/main/think.py
https://gist.github.com/idvorkin/25eadce2222a09d9dca620679afd59d6
Other folks work here
- A different version of think.py, from danielmiessler lots of much better prompts