Notes
A growing web of short, linked ideas about augmented intelligence — the conviction that AI is one tool for thinking, not the thinker. Each note states one claim and links to the others. Follow a thread; you won't lose your place.
How to read this: these are evergreen notes — each titled with a single claim, kept short, and linked to its neighbours. Start anywhere and follow the links. "Linked from" at the foot of each note shows what points back to it.
- A good explanation is hard to varyA real explanation commits to specifics you can't swap out without breaking it; a fake one would fit any outcome equally well.
- A page can remember what you learnProse interleaved with spaced recall prompts turns ordinary reading into remembering.
- AI is one tool for thinking, not the thinkerAI is the newest tool for extending thought, but it never supplies the goal, the judgment, or the responsibility.
- Augmenting a person beats automating the taskAutomation caps quality at the system's blind spots; augmentation keeps the human in the loop and raises the ceiling.
- Autonomy is not the same as agencyAutonomy is freedom from interference; agency is the capacity to set and pursue your own goals.
- Delegate the task, keep the judgmentYou can hand an AI the doing without handing over the deciding about what matters and whose interest wins.
- Evergreen notes are written to evolve, not to fileMost notes are written once and abandoned; evergreen notes are atomic and rewritten over time, so they accumulate instead of rot.
- Knowledge compounds when it's linked, not just storedA pile of saved notes is inert; the value lives in the links between ideas that make one note reach another.
- Retrieval, not rereading, is what makes it stickPulling an idea back out of your head strengthens memory far more than reading it again.
- Show the data; don't just assert itUnderstanding comes from seeing and poking the real thing, not from reading a claim someone made about it.
- Spacing the practice fights the forgetting curveMemories decay predictably; recalling just before you would forget resets the clock each time.
- Structure beats freeform text for working with AIA defined shape makes an AI request precise and its output checkable field by field; freeform prose is fluent but unverifiable.
- The interface shapes what you can thinkThe same AI behind a chat box, a canvas, or a direct tool produces very different thinking.
- The skill compounds, so start nowWorking well with AI is a skill built from reps, and skills compound, so a small head start widens over time.
- Tools for thought change what you can thinkA new way to represent ideas does not just record thinking; it makes new thoughts reachable.
- Trust is a calibrated forecast, not a feelingTrusting an AI output should be a probability you update, not a vibe from how fluent it sounds.