Papers I Keep Rereading

Diego Navarro

@turbochargediego

8Items

0Likes

When I’m not doomscrolling Heat Twitter or arguing about product roadmaps, I’m usually elbow‑deep in PDFs. This is the little canon I keep coming back to — the stuff that actually changed how I ship features, argue in meetings, and generally make sense of people plus software (plus my own bad decisions). It’s a weird mix: some HCI, some behavioral econ, some ML bangers. Not a “read in order” syllabus — more like a late‑night queue you dip into when your brain wants something denser than a blog post but friendlier than a textbook.

See all 8 items Follow Diego NavarroOpen in app

The list

Showing 1–8 of 8

Papers I Keep Rereading — 8 items
Name	Field	Year	Main topic	Key idea	Why it matters	Best for	Difficulty	My take	Reading tip
1The Magical Number Seven, Plus or Minus Two	Cognitive psychology	1956	Limits of human short-term memory	People can hold about 7 (give or take) discrete chunks in working memory at once.	Explains why overloaded menus, giant forms, and 20-option settings pages feel like getting full-court pressed.	Pruning UI complexity, onboarding flows, and settings architecture.	Medium — old-school academic prose but short.	Read this and you’ll start seeing “magic seven” landmines all over your product — and in your slide decks.	Skim the intro, then focus on the sections with examples of chunking; you can mostly ignore the mathy bits.
2Prospect Theory: An Analysis of Decision under Risk	Behavioral economics	1979	How people actually make risky choices	We hate losses more than we like equivalent gains, and we’re irrationally weird about probabilities.	Frames discounts, fees, and upgrade prompts in terms of perceived loss vs gain — huge for pricing and retention.	Designing plans, paywalls, and nudges that don’t backfire.	High — dense, with lots of formalism, but the intuition is gold.	If you’ve ever wondered why users rage-quit over a “small” fee, this paper is the scouting report.	Read a summary first, then dive into the value function and loss aversion sections with a notebook.
3The Strength of Weak Ties	Sociology / networks	1973	How social networks actually move information	Loose connections (weak ties) are disproportionately powerful for spreading information and opportunities.	Explains why features around acquaintances, not just close friends, can drive discovery and growth.	Social features, referral loops, marketplace design.	Medium — clear writing, some stats but nothing wild.	Makes you rethink “friends list” UX; the casual contacts are doing more work than your besties.	Focus on the early theory sections; you can skim the deeply technical bits on empirical methodology.
4The Cathedral and the Bazaar	Software engineering / open source	1997	Contrasting centralized vs open development	Given enough eyeballs, bugs are shallow — open, iterative collaboration can beat top-down planning.	Foundational for how we think about open source, but also eerily relevant to modern product and community building.	Thinking about contribution models, community roadmaps, and platform ecosystems.	Low — more essay than formal paper.	Reads like the origin story for GitHub, hackathons, and half of modern dev culture.	Print and mark up; it’s short enough to read in one sitting and argue with in the margins.
5Attention Is All You Need	Machine learning / NLP	2017	Transformer architecture for sequence modeling	Dumps recurrence and convolutions and leans on self-attention to model long-range dependencies efficiently.	Basically the blueprint for modern language models; understanding this changes how you think about AI features.	Anyone spec’ing AI features or trying to talk sanely with ML engineers.	High — mathematical, but the diagrams are friendly.	This is the playbook behind half the AI hype decks; worth reading the actual play instead of just the commentary.	Read the intro and architecture overview, then jump straight to the diagrams and skip most derivations on first pass.
6Do Defaults Save Lives?	Behavioral economics / policy	2003	Power of defaults in organ donation choices	Simply setting a default massively changes participation rates without changing incentives or information.	Makes you treat every toggle and pre-checked box like a loaded weapon — defaults steer behavior hard.	Designing settings, privacy controls, and onboarding choices.	Low to medium — very readable.	Once you see their charts, you’ll never casually pick a default state again; it’s pure behavioral leverage.	Read the results and discussion sections slowly; mentally substitute your product’s key opt-ins.
7The Spread of True and False News Online	Computational social science	2018	How misinformation propagates on social platforms	False news spreads faster, farther, and deeper than true news, largely driven by humans, not bots.	If you touch feeds, notifications, or sharing, this is the cautionary tale you should have in your head.	Feed ranking, moderation policy, and virality mechanics.	Medium — data-heavy but well explained.	Reads like a postgame breakdown of a blowout loss where the defense (us) never adjusted to the fast break (misinfo).	Focus on the main figures and their explanations; you don’t need every methodological detail to get the story.
8Designing for Homo Heuristicus	Behavioral economics / design	2011	Designing systems for biased, shortcut-using humans	People rely on simple heuristics; good systems align with those heuristics instead of fighting them.	Great mental model for why “rational user” assumptions keep breaking your flows in the wild.	Choice architecture, dashboards, and anything with lots of options.	Medium — conceptual but not super mathy.	This is the paper that made me stop designing for Ideal Me and start designing for Sleepy 11:30 p.m. Me.	Read with a specific product in mind; after each heuristic, jot how your UI currently trips over it.