Spaced repetition appears 'accidental' in many professional knowledge practices: writing background sections, mentoring conversations, medical rounds. The practices achieve spaced repetition effects naturally without explicit intent.

Misery in learning comes primarily from resisting struggling—feeling like you're not progressing fast enough or failing, believing you shouldn't struggle, fearing judgment—rather than from the inherent difficulty of the material itself.

Spreading activation theory explains forgetting as a functional feature—if you didn't forget, you'd have a cacophony of noise in memory retrieval; forgetting helps you focus on what's salient and relevant to the current situation by tuning weights on memory connections.

'Cool shit' in innovation is likely to come from disaffected and bored high-school students (20th percentile performers) who are not engaged by traditional education, suggesting that by not helping these students learn, society loses potential innovations.

Matuschak endorses the traditional view that taxonomy memorization is important (following a 1900s headmaster's view), but only because modern memory systems make it trivial in cost—the value proposition changes fundamentally when memorization takes 15 minutes total rather than months.

Crowdfunded research faces churn: ~2% monthly subscriber loss requires constant marketing to maintain. Beyond initial interest, churn is surprisingly insensitive to content quality or frequency of updates.

At Apple, Matuschak was much better at engineering work than he is now at research work, because engineering drew on deeply familiar knowledge in a well-understood domain requiring flow and execution, while research requires constantly answering 'what should be done' questions that demand synthesizing unfamiliar literature.

Streaming (like George Hotz streaming code) can transmit tacit knowledge—things not in books that engineers learn through long experience dealing with unusual situations—and people can absorb pattern recognition and heuristics by watching, but streaming lacks feedback and contains lots of chaff.

Memory is underappreciated in learning and life; your ability to understand difficult material is absolutely bound by your memory of the constituent elements, because understanding complex arguments requires holding and manipulating multiple pieces in relation to each other.

Matuschak hasn't published 'minimum viable papers' (marginal insights above publication threshold) even though this costs him financially. This reflects commitment to depth over productivity.

Paul Graham's claim that reading trains your model of the world and those effects persist even if you forget the specific content is true to some extent, but whether that extent is sufficient depends heavily on what you need—if your aspiration requires deep detailed understanding, then just imprinting on your worldview may not be enough.

The education system for the median student is primarily structured around making students do things they don't want to do and achieve goals that aren't theirs, rather than helping them achieve their own goals more effectively.

Active reading—asking questions of a text, probing what you don't understand, and dwelling on confusion—is a learnable skill that dramatically improves comprehension, and most readers fail not because they forget but because they never notice they didn't understand in the first place.

Scaffolding in education—providing simpler versions of problems first, partially worked solutions, or worked examples—is effective for building capability, but should be faded (gradually removed) as learners become more capable, because continuing scaffolding becomes ineffective as learners progress.

At Apple, product design and strategy involves compartmentalized domains with clear boundaries—each leader has a small 5% zone of direct control, a surrounding ring of things they monitor and advise on, and delegated domains they only hear about if there are problems.

Quality at a certain level (like Gwern's research) probably doesn't need traditional marketing because high-quality work spreads naturally, but Matuschak's experience is that many people don't know he has a Patreon despite being familiar with his work—suggesting some marketing of membership offerings (different from promoting content) is efficient.

Creative insight often depends on memory: noticing surprising connections between disparate ideas requires those ideas to be in memory and salient; if you must consciously retrieve an idea, it won't seem salient when you encounter related information, preventing the creative leap.

ACT-R theory by John Anderson explains learning through 'knowledge compilation'—the process of converting individual facts into higher-level patterns that can be generalized and applied in new contexts—and Anderson's theory suggests this compilation requires repeated exposure, active use, and variety of contexts, not just reading.

Dwarkesh's interviewing skill has improved over time not because he remembers details from past conversations, but because he's developed automatic patterns for generating good questions about material—through repeated practice reading essays and blog posts, these question-generation patterns have become automatic through knowledge compilation.

Content infovores like Tyler Cowen integrate large amounts of information without formal note-taking because their writing obligations (daily blog posts, weekly columns, frequent books) function as distributed note-taking—the act of writing down their thinking, often in first-draft form, is how they process and integrate material.