Recommended Reading from Tech Founders

Strongest signal: the Sam Altman essay Garry Tan says "opened my eyes"

Among today’s items, this is the clearest recommendation tied to a change in thinking. Tan says a Sam Altman essay he refers to as Age of Intelligence "really opened my eyes" to what he thinks builders should do next SXSW.

• Title:Age of Intelligence (as Garry Tan names it) SXSW
• Content type: Essay SXSW
• Author/creator: Sam Altman SXSW
• Who recommended it: Garry Tan SXSW
• Key takeaway: Tan says it changed how he thinks about what to do from here and pairs that with criticism of what he sees as too much doomerism from some frontier labs SXSW
• Why it matters: This is the strongest signal in the batch because the endorsement is explicitly about changed thinking, not generic praise SXSW

Same recommender, much shorter format

Tan also reaches for a very different kind of resource: a compact book he describes as simple but foundational SXSW.

• Title:Who Moved My CheeseSXSW
• Content type: Book SXSW
• Author/creator:Not specified in the provided material
• Who recommended it: Garry Tan SXSW
• Key takeaway: He calls it "a short, very simple book" and also "the defining" one SXSW
• Why it matters: The strength of the endorsement stands out given how compact he says the book is SXSW

Best map for a crowded AI topic

Scott Belsky’s most useful recommendation today is an explanatory X thread by @zhuokaiz that he calls "good posts on blurring lines of varied approaches to so-called world models…" @scottbelsky

• Title:Five categories of world models@zhuokaiz
• Content type: X thread @scottbelsky
• Author/creator: @zhuokaiz @scottbelsky
• Link/URL:https://x.com/zhuokaiz/status/2032201769053212682@scottbelsky
• Who recommended it: Scott Belsky @scottbelsky
• Key takeaway: The thread is useful because it does not flatten "world models" into one idea; it organizes the space into JEPA, spatial intelligence, learned simulation, NVIDIA Cosmos, and active inference @zhuokaiz
• Why it matters: For readers trying to get oriented quickly, a five-part taxonomy is more actionable than treating the whole field as a single bucket @zhuokaiz

A few concrete anchors from the thread:

• JEPA / V-JEPA 2: latent-space prediction instead of pixel reconstruction; after large-scale video pretraining, just 62 hours of robot data is described as enough for zero-shot planning @zhuokaiz
• Spatial intelligence / Marble: persistent 3D environments that can be generated from images, text, video, or 3D layouts @zhuokaiz
• Learned simulation: the thread argues generative video models and RL world models are converging around the same need—simulating how actions change environments over longer horizons @zhuokaiz
• NVIDIA Cosmos: positioned as a platform play spanning data curation, tokenization, training, and deployment rather than one world model alone @zhuokaiz
• Active inference / AXIOM: an object-centric, Bayesian alternative to monolithic neural world models, with robotics examples built around hierarchical agents and online inference @zhuokaiz

One offbeat repeat-view pick

This is the outlier in today’s set, but it is still a clear organic recommendation because Graham emphasizes repeat viewing, not novelty @paulg.

• Title:The Larry Sanders Show@paulg
• Content type: Show / series @paulg
• Author/creator:Not specified in the provided material
• Who recommended it: Paul Graham @paulg
• Key takeaway: He says the show is "so amazing," says he is watching it again "for about the fourth time," and calls it "brilliant people skewering a world they know all too well" @paulg
• Why it matters: The repeat-viewing detail makes this feel more durable than a one-off mention @paulg

What stands out

The most useful recommendations today do one of two things: they either clearly change a leader’s posture or they make a crowded area legible. Garry Tan’s Sam Altman essay recommendation is strong because he explicitly says it reframed what to do next, while Scott Belsky’s thread recommendation is useful because it gives readers a compact map of competing world-model approaches SXSW +2.

Strongest signal: Beyond Belief

In the provided material, this is the clearest recommendation because it shows repeat engagement, not a one-off mention. Ryan Hoover said Nir Eyal had announced his new book, Beyond Belief@rrhoover, noted that he helped Eyal write Hooked 13 years ago and "learned way more than I contributed" @rrhoover, then said he pre-ordered the audiobook @rrhoover and later that he was listening once it released @rrhoover.

• Title:Beyond Belief@rrhoover
• Content type: Book / audiobook @rrhoover +1
• Author/creator: Nir Eyal @rrhoover
• Link/URL: Recommendation source posts: announcement, audiobook release
• Who recommended it: Ryan Hoover @rrhoover +1
• Key takeaway: The source material does not surface a specific lesson from the book itself. The usable signal is Hoover's follow-through from announcement to pre-order to release-day listening @rrhoover +1
• Why it matters: Hoover brings relevant context to the recommendation because he previously helped Eyal write Hooked and says he learned more than he contributed @rrhoover

Another authentic pick: The Internet Is Gonna End Us (But it’s okay!)

Marc Andreessen highlighted HeavyPulp’s video by linking directly to it and calling it the "second best thing I’ve ever seen" @pmarca +1.

• Title:The Internet Is Gonna End Us (But it’s okay!)@heavypulp
• Content type: Video @heavypulp
• Author/creator: @heavypulp @pmarca
• Link/URL:https://x.com/heavypulp/status/2015921562038206851@pmarca
• Who recommended it: Marc Andreessen @pmarca
• Key takeaway: The source material does not include a substantive lesson from the video itself; the main signal is the intensity of Andreessen’s endorsement @pmarca
• Why it matters: Even without added explanation, this stands out as a direct and unusually strong recommendation tied to a specific piece of content rather than a vague reference @pmarca

What to take from this set

The pattern here is strength of endorsement over articulated lesson. Beyond Belief is the more useful signal because the recommendation is repeated and grounded in Hoover’s prior working relationship with Eyal @rrhoover +1. Andreessen’s video share is still notable, but the provided material says much less about what viewers should expect beyond his enthusiasm @pmarca.

Most compelling recommendation: Xiaoyu Ma and David Patterson on AI inference hardware

This stands out because the endorsement comes with a precise diagnosis and a concrete design agenda. Chamath frames the next AI silicon cycle around "cheap, abundant decode," while the underlying paper argues that inference, especially decode, is constrained by memory bandwidth and memory cost more than raw compute @chamath +1.

• Title: IEEE Computer 2026 paper by Xiaoyu Ma and David Patterson on AI inference hardware (official title not provided in the source material)@ChrisLaubAI
• Content type: Research paper @chamath +1
• Author/creator: Xiaoyu Ma and David Patterson @ChrisLaubAI
• Link/URL: Discussion post: https://x.com/ChrisLaubAI/status/2032035780189962292
• Who recommended it: Chamath Palihapitiya @chamath
• Key takeaway: GPU FLOPS have outpaced memory bandwidth, HBM costs per GB are rising, and decode is memory-bound; the paper's proposed shifts include high-bandwidth flash, processing-near-memory, 3D memory-logic stacking, and lower-latency interconnects @ChrisLaubAI
• Why it matters: The recommendation is paired with a clear thesis about what gets worse from here: MoE models, reasoning chains, multimodal inputs, long context windows, and RAG all increase pressure on inference hardware @ChrisLaubAI

Founder and operator playbooks

Long Strange Trip episode with Kaz Nejatian

• Title:Long Strange Trip episode with Kaz Nejatian of Opendoor (episode title not provided in the source material)@bhalligan
• Content type: Podcast episode @rabois +1
• Author/creator: Brian Halligan / Long Strange Trip@bhalligan
• Link/URL:https://x.com/bhalligan/status/2032132659384840642@rabois
• Who recommended it: Keith Rabois @rabois
• Key takeaway: Halligan's summary turns the conversation into a compact founder playbook: build "first derivative" businesses, reject inherited defaults, optimize for stewardship over status, write a user manual for yourself, and hold yourself responsible for outcomes rather than process @bhalligan
• Why it matters: The lessons are framed in the context of Kaz Nejatian helping refound a struggling public company in 16 days, giving the abstractions a concrete operating backdrop @bhalligan

The Founder's Dilemmas

• Title:The Founder's DilemmasHarvey Mudd College
• Content type: Book Harvey Mudd College
• Author/creator: Noam Wasserman Harvey Mudd College
• Link/URL: Source conversation: https://www.youtube.com/watch?v=ZuMctpPgTms
• Who recommended it: Josh Jones Harvey Mudd College
• Key takeaway: Jones highlights the book's distinction between founders trying to "get rich" and founders trying to "be their own boss," and says co-founder teams can break when those motivations do not match Harvey Mudd College
• Why it matters: He describes the book as the product of long-run founder survey data and says he wished he had read it before starting his first company Harvey Mudd College

Kadampa Meditation

• Title:Kadampa MeditationHarvey Mudd College
• Content type: Book Harvey Mudd College
• Author/creator: Geshe Kelsang Gyatso Harvey Mudd College
• Link/URL: Source conversation: https://www.youtube.com/watch?v=ZuMctpPgTms
• Who recommended it: Nihal Mehta Harvey Mudd College
• Key takeaway: Mehta describes the practice as pursuing liberation through the alleviation of human suffering, and says he applies that idea to venture by trying to help people reach their potential Harvey Mudd College
• Why it matters: It is a rare recommendation that explicitly connects inner practice to how an investor thinks about work, service, and happiness Harvey Mudd College

Why today's list is useful

Across very different formats, the common thread is clearer diagnosis. The paper asks what is actually bottlenecking AI systems, the podcast asks what leaders should actually own, The Founder's Dilemmas asks what founders are actually optimizing for, and Kadampa Meditation asks what kind of service sits underneath ambition @ChrisLaubAI +2

Most compelling recommendation: Paul Graham’s essays

This stands out because the endorsement is both repeated and concrete. Matt Mullenweg says he returns to Paul Graham’s writing "again and again" @photomatt, and the thread he resurfaced says the essays have served as a "business coach" for the last 10 years and are reread annually when product or team questions come up @tibo_maker +1.

• Title: Paul Graham’s essays
• Content type: Essays / blog
• Author/creator: Paul Graham @photomatt
• Link/URL: Source thread with the reread list: https://x.com/tibo_maker/status/2031679065099284971@photomatt
• Who recommended it: Matt Mullenweg, who said he returns to Paul Graham’s writing repeatedly and highlighted a thread from @tibo_maker detailing the essays he rereads yearly @photomatt +1
• Key takeaway: In the highlighted thread, the essays are treated as practical startup education and a reusable operating manual for customer development, idea generation, time management, founder involvement, and value creation @tibo_maker
• Why it matters: The source material ties the recommendation to concrete operating decisions: acting as first-line support, protecting maker time, using launch checklists, staying close to details, and focusing on creating value rather than extracting it @tibo_maker

The clearest lessons surfaced in the thread:

• Do things that don’t scale — founders should do manual work early; the poster says he still acts as the first customer support rep because that teaches more than dashboards @tibo_maker
• How to get startup ideas — the best ideas come from living at the edge of a problem, not from brainstorming market size @tibo_maker
• Maker’s schedule, manager’s schedule — makers need uninterrupted half-day blocks; the applied version here is no meetings and async Slack @tibo_maker
• Founder mode — standard delegation advice can be the wrong playbook for founders who need to stay close to details @tibo_maker
• How to make wealth — startups should create new value for people rather than extract it @tibo_maker
• Life is short / How to do great work — the broader frame is finite time and work that sits at the intersection of natural ability, obsession, and ambition @tibo_maker

Other durable picks

Sparks of Artificial General Intelligence (2023)

• Title:Sparks of Artificial General Intelligence (2023)
• Content type: Research paper
• Author/creator: Sébastien Bubeck @pmarca
• Who recommended it: Marc Andreessen @pmarca
• Key takeaway: Andreessen’s recommendation is concise but clear: the paper is "aging very well" @pmarca
• Why it matters: In a fast-moving AI cycle, the signal here is durability: he is pointing readers back to a 2023 paper as still worth attention now @pmarca

WarGames (1983)

• Title:WarGames (1983)
• Content type: Film
• Author/creator: Not specified in the source material
• Who recommended it: Reid Hoffman Reid Hoffman
• Key takeaway: Hoffman uses the film as an example of an AI reasoning through "no win" scenarios and learning that escalation is not the right move Reid Hoffman
• Why it matters: He presents it as a case for keeping humans in the loop, emphasizing that pure rationality needs compassion, context, and judgment—including moments when people refuse to trust what sensors appear to show Reid Hoffman

Pattern across today’s recommendations

The shared trait is staying power: a paper that still holds up years later, an essay corpus reread as an operating manual, and a 1983 film still used to think through AI risk and human oversight @pmarca +2

Most compelling recommendation: Apple in China

This stands out because it comes with a concrete mechanism, not just praise. Shyam Sankar recommends Patrick McGee's Apple in China to argue that manufacturing talent and industrial capacity are built through sustained investment, citing Apple's inflation-adjusted spending in China as his evidence Invest Like The Best.

• Content type: Book Invest Like The Best
• Author/creator: Patrick McGee Invest Like The Best
• Who recommended it: Shyam Sankar Invest Like The Best
• Key takeaway: Sankar uses the book to rebut claims that the U.S. simply lacks machine-tool engineers; his point is that Apple invested heavily to build talent and capacity in China Invest Like The Best
• Why it matters: It turns a broad debate about competitiveness into a specific case study of how capability gets created Invest Like The Best

Builder and strategy reads

Rickover's personal memoirs

• Content type: Book Invest Like The Best
• Author/creator: Hyman Rickover Invest Like The Best
• Who recommended it: Shyam Sankar Invest Like The Best
• Key takeaway: Sankar points readers to Rickover's memoirs to see how he channeled humiliation into motivation while pushing ahead on nuclear submarines despite opposition Invest Like The Best
• Why it matters: It is a firsthand builder account of how one difficult operator handled institutional resistance Invest Like The Best

Biography of John Boyd

• Content type: Biography Invest Like The Best
• Author/creator: Not specified in the source material Invest Like The Best
• Who recommended it: Patrick O'Shaughnessy Invest Like The Best
• Key takeaway: Patrick describes it as one of the great biographies of a military figure he has read and highlights Boyd's invention of the OODA Loop Invest Like The Best
• Why it matters: It pairs a life story with a named decision framework Patrick explicitly calls out Invest Like The Best

Destined for War with China

• Content type: Book Invest Like The Best
• Author/creator: Graham Allison Invest Like The Best
• Who recommended it: Patrick O'Shaughnessy Invest Like The Best
• Key takeaway: Patrick says he reviewed notes from the book before the conversation to frame how he thinks about China as an adversary and as a country Invest Like The Best
• Why it matters: It is presented as active preparation for a current geopolitical discussion, not as a generic reading-list item Invest Like The Best

Zach Dell's learning stack

Zach Dell surfaced four resources in the materials for his Generalist interview The Generalist. The source material gives less commentary on these picks, so the signal is mainly in the shape of the list itself.

Factfulness: Ten Reasons We're Wrong About the World--and Why Things Are Better Than You Think

• Content type: Book The Generalist
• Author/creator: Not specified in the source material The Generalist
• Link/URL:amazon.com/Factfulness-Reasons-World-Things-Better/dp/1250107814The Generalist
• Who recommended it: Zach Dell The Generalist
• Key takeaway: The title frames it as a corrective to common misreadings of the world The Generalist
• Why it matters: Within Dell's four picks, it is the broadest worldview recommendation The Generalist

Robert Greene's books

• Content type: Books / author body of work The Generalist
• Author/creator: Robert Greene The Generalist
• Link/URL:amazon.com/stores/Robert-Greene/author/B001IGV3ISThe Generalist
• Who recommended it: Zach Dell The Generalist
• Key takeaway: Dell recommends Greene's work as a body of books rather than pointing to one specific title The Generalist
• Why it matters: It is the only pick in today's set framed as an author's broader catalog The Generalist

Childhoods of exceptional people

• Content type: Article / blog post The Generalist
• Author/creator: Not specified in the source material The Generalist
• Link/URL:henrikkarlsson.xyz/p/childhoodsThe Generalist
• Who recommended it: Zach Dell The Generalist
• Key takeaway: The piece focuses on the childhoods of exceptional people The Generalist
• Why it matters: It adds a development-focused lens alongside worldview and learning-method reads The Generalist

Feynman Technique: The Ultimate Guide to Learning Anything Faster

• Content type: Article The Generalist
• Author/creator: Not specified in the source material The Generalist
• Link/URL:fs.blog/feynman-techniqueThe Generalist
• Who recommended it: Zach Dell The Generalist
• Key takeaway: It is a direct recommendation on learning anything faster through the Feynman Technique The Generalist
• Why it matters: It is Dell's clearest explicitly process-oriented learning resource The Generalist

Pattern across today's recommendations

No single resource was repeated by multiple leaders, but the day's picks cluster cleanly. Sankar's recommendations focus on industrial and military builders, Patrick's on strategy and China, and Dell's on learning, worldview, and the development of exceptional people Invest Like The Best +1.

Most compelling recommendation: autoresearch

This stands out because it comes with both a concrete operating model and a reported result. Andrej Karpathy’s autoresearch packages a minimal, single-GPU LLM training core into a self-contained repo; Tobi Lütke says adapting the approach to his qmd query-expansion model produced a +19% score on a 0.8B model, above his previous 1.6B model, after 37 experiments in 8 hours @karpathy +1.

autoresearch

• Content type: GitHub repo / code resource
• Author/creator: Andrej Karpathy @karpathy
• Link/URL:github.com/karpathy/autoresearch@karpathy
• Who recommended it: Tobi Lütke, who shared his results after adapting the setup to his own model work @tobi
• Key takeaway: The human iterates on the prompt while the AI agent iterates on the training code in an autonomous git loop, searching for better architectures, optimizers, and hyperparameters via repeated short training runs @karpathy
• Why it matters: Lütke says the setup beat his previous baseline on a smaller model, already built a better reranker, and was valuable as a learning tool in its own right @tobi

Qasar Younis’s founder reading stack

Younis’s strongest meta-recommendation is to read old, time-filtered books across history, business, and society rather than low-quality new content. He argues that broader reading makes founders more well rounded and helps them build better products Lenny's Podcast.

The Emperor of All Maladies

• Content type: Book
• Author/creator: Siddhartha Mukherjee Lenny's Podcast
• Who recommended it: Qasar Younis Lenny's Podcast
• Key takeaway: He says it changes how you think, and treats that shift in framing as the test of great material Lenny's Podcast
• Why it matters: It shows the kind of non-tech book Younis values: one that changes your worldview rather than offering a direct startup tactic Lenny's Podcast

SPQR

• Content type: Book
• Author/creator: Mary Beard Lenny's Podcast
• Who recommended it: Qasar Younis Lenny's Podcast
• Key takeaway: He picked it after realizing he did not know Roman history well and uses books like this to fill important gaps in knowledge Lenny's Podcast
• Why it matters: It turns his broader heuristic into something actionable: identify an area you do not understand, then read the best book in that space Lenny's Podcast

Made in America

• Content type: Book
• Author/creator: Sam Walton Lenny's Podcast
• Who recommended it: Qasar Younis Lenny's Podcast
• Key takeaway: Younis calls it an unbelievable book and notes that Walton wrote it on his deathbed Lenny's Podcast
• Why it matters: It is one of the durable business histories he elevates over low-quality content, consistent with his view that founders should spend scarce reading time on books that have already survived time’s filter Lenny's Podcast

House of Huawei

• Content type: Book
• Author/creator: Not specified in the source material
• Who recommended it: Qasar Younis, who says his company recently read it together Lenny's Podcast
• Key takeaway: He calls it a really great, interesting book and pairs it with admiration for Huawei as a company that makes great technology Lenny's Podcast
• Why it matters: It was strong enough to become a company read, making it a useful organizational case study rather than just a personal pick Lenny's Podcast

Guns, Germs, and Steel / Collapse

• Content type: Books
• Author/creator: Jared Diamond Lenny's Podcast
• Who recommended them: Qasar Younis Lenny's Podcast
• Key takeaway: He places both near the top of his list and calls them fantastic Lenny's Podcast
• Why it matters: They are clear examples of the non-tech, time-tested books he thinks founders should choose over lower-quality content Lenny's Podcast

Two worldview reads on intelligence and the internet

What is Intelligence?

• Content type: Book
• Author/creator: Not specified in the source material
• Link/URL:mitpress.mit.edu/9780262049955/what-is-intelligence/@tobi
• Who recommended it: Tobi Lütke @tobi
• Key takeaway: Lütke calls it a great book while agreeing with the claim that computer science is becoming an explanation of how the world works, like physics, biology, and chemistry @tobi +1
• Why it matters: The recommendation is attached to a broad framing of computing as foundational knowledge, not just a useful craft @tobi +1

A Declaration of the Independence of Cyberspace

• Content type: Manifesto / article
• Author/creator: John Perry Barlow @photomatt
• Link/URL:eff.org/cyberspace-independence@photomatt
• Who recommended it: Matt Mullenweg @photomatt
• Key takeaway: Mullenweg says it is an apt time to revisit the text roughly 30 years later and praises its poetry @photomatt
• Why it matters: It is an explicit recommendation to go back to an older primary text, not just commentary about the internet @photomatt

Pattern across today’s recommendations

The common thread is a bias toward foundations over novelty: Younis tells founders to read old books, Mullenweg revives a 30-year-old internet text, Lütke links intelligence to computing as a basic explanatory lens, and his strongest practical pick is a stripped-down repo that exposes the mechanics of model research Lenny's Podcast +3.

Most compelling recommendation: product learning before automation

This stands out because Garry Tan does more than restate a classic startup essay. He explains why its core lesson becomes more important when AI makes premature scaling easy @garrytan.

Do Things That Don’t Scale — Paul Graham

• Content type: Essay
• Author/creator: Paul Graham
• Link/URL: Not included in the source material
• Recommended by: Garry Tan
• Recommendation context:Garry Tan on X
• Key takeaway: Tan says the essay still matters because its real point is contact with users: manual work reveals edge cases, weird behavior, failure modes, hidden dependencies, and the actual jobs to be done before you automate @garrytan.
• Why it matters: In Tan’s framing, good abstractions are discovered by doing the manual loop until the real bottleneck emerges, not designed upfront from a hunch @garrytan.

Another useful recommendation: Howard Marks’ market-discipline heuristic

Howard Marks’ investor letters — Howard Marks

• Content type: Investor letters / memos
• Author/creator: Howard Marks
• Link/URL: Not included in the source material
• Recommended by: Harry Stebbings
• Recommendation context:20VC episode
• Key takeaway: Stebbings highlights one Marks principle in particular: never try to “catch a falling knife” when assessing buying opportunities during market declines 20VC with Harry Stebbings.
• Why it matters: It is a compact rule for staying disciplined in downturns, and Stebbings presents Marks’ letters as a recurring source of that judgment 20VC with Harry Stebbings.

Pattern across today’s recommendations

Both picks warn against premature action. Tan’s recommendation is about not scaling before direct contact shows what users actually need; Stebbings’ recommendation is about not rushing into a falling market too early @garrytan +1.

Most compelling recommendation: a crisp mental model for what LLMs optimize vs. what they struggle to do

Shannon Got AI This Far, Kolmogorov Shows Where It Stops — Vishal Misra (Medium article)

• Content type: Article (Medium)
• Author/creator: Vishal Misra
• Link/URL:https://medium.com/@vishalmisra/shannon-got-ai-this-far-kolmogorov-shows-where-it-stops-c81825f89ca0@vishalmisra
• Recommended by: Martin Casado (@martin_casado)
• Key takeaway (as shared): Casado calls it a “great analogy describing what LLMs can and can’t do,” framing them as:
  • Good at cross-entropy loss (predicting what’s next in training data) @martin_casado
  • Bad at reducing Kolmogorov complexity (finding a dramatically simpler underlying program/solution that generates the data) @martin_casado
• Why it matters: This is a compact, transferable lens for evaluating where LLMs may excel (next-token-style prediction) vs. where they may fall short (discovering radically simpler generative explanations), without requiring a long debate about “intelligence.” @martin_casado

Product + media: don’t retrofit—redesign the broadcast

Apple, You Still Don’t Understand the Vision Pro — Ben Thompson (Stratechery article)

• Content type: Article (Stratechery)
• Author/creator: Ben Thompson
• Link/URL:https://stratechery.com/2026/apple-you-still-dont-understand-the-vision-pro/@garrytan
• Recommended by: Garry Tan (@garrytan)
• Key takeaway (as shared): Tan agrees with Stratechery that the Apple Vision Pro product team should be thinking about remaking sports broadcasting from the bottom up—and contrasts that with a Lakers game broadcast that sounded like “building a faster horse.” @garrytan
• Why it matters: It’s a clear “blank-sheet” product prompt: if you’re building in a new medium (spatial computing), the opportunity may be to rebuild the experience, not just port existing formats with incremental upgrades. @garrytan

Policy/economics reading flagged by founders

SEIU delenda est — Astral Codex Ten (article)

• Content type: Article
• Author/creator: Astral Codex Ten
• Link/URL:https://www.astralcodexten.com/p/seiu-delenda-est@paulg
• Recommended by: Paul Graham (@paulg)
• Key takeaway (as shared): Graham says the article explains clearly why California’s proposed wealth tax would be damaging, and adds that it’s “not an accident” but “designed to be damaging.” @paulg
• Why it matters: If you track California policy and its downstream effects on startups/founders, this is a pointed “read this to understand the argument” recommendation from a high-signal operator. @paulg

Pattern across today’s picks

A common thread is first-principles framing: one recommendation offers a clean boundary model for LLM capabilities (cross-entropy vs. Kolmogorov complexity) @martin_casado, another argues for redesigning an experience “from bottom up” rather than incremental upgrades @garrytan, and a third spotlights an argument about policy incentives and intended effects @paulg.

Most compelling recommendation: a check on “improvement” becoming its own addiction

The Self-Help Trap: What 20+ Years of “Optimizing” Has Taught Me — Tim Ferriss (blog post)

• Type: Blog post @tferriss
• Author/creator: Tim Ferriss @tferriss
• Link/URL:https://x.com/tferriss/status/2029283224866770944@ShaanVP
• Recommended by: Shaan Puri (@ShaanVP) @ShaanVP
• Key takeaway (as shared): Shaan connects Ferriss’ point to seeing people at a Tony Robbins event who, after attending 3+ times, seemed to get “addicted to the medicine” @ShaanVP.
• Why it matters: It’s a rare, high-conviction endorsement (“my favorite thing Tim has written in 10+ years”) aimed directly at “fellow self improvers” @ShaanVP—useful if you’re trying to ensure learning and self-improvement translate into changed behavior (not just repeated consumption).

Engineering fundamentals worth (re)loading into your brain

Video on database consistency + concurrency tradeoffs — @jamwt (video)

• Type: Video (posted on X) @jamwt
• Author/creator: @jamwt @jamwt
• Link/URL:https://x.com/jamwt/status/2029353984792961278@martin_casado
• Recommended by: Martin Casado (@martin_casado) @martin_casado
• Key takeaway (as shared): A “fantastic overview” demystifying database consistency, isolation levels, record contention, and pessimistic vs. optimistic concurrency control tradeoffs@martin_casado +1.
• Why it matters: Casado frames these as “super important concepts” if you’re “building production systems,” and suggests that as AI reduces the need to “memorize random framework nonsense,” this is the kind of broadly useful material to replace it with @martin_casado.

AI is changing workflows (and the signal-to-noise ratio)

The Tinder-ization of the Job Market — Matt Darling (essay)

• Type: Article/essay Not Boring by Packy McCormick
• Author/creator: Matt Darling (The Argument) Not Boring by Packy McCormick
• Link/URL:https://www.theargumentmag.com/p/the-tinder-ization-of-the-job-marketNot Boring by Packy McCormick
• Recommended by: Packy McCormick (Not Boring) Not Boring by Packy McCormick
• Key takeaway (as highlighted): The argument presented is that the job market is “stuck” (e.g., hiring rate averaged 3.3% in H2 2025) even while unemployment was 4.3% in January and prime-age (25–54) employment was 80.9%Not Boring by Packy McCormick. One proposed mechanism: LLMs make it easier to apply to many jobs, increasing volume while weakening traditional signals (e.g., recruiting workload rose 26% in Q3 2024; 38% of job seekers reported “mass applying”; an applications-to-recruiter ratio “about 500–1”) Not Boring by Packy McCormick.
• Why it matters: If you hire, this is a concrete pointer to why screening may be breaking down under application flooding and AI-generated materials—and why process adjustments may be required Not Boring by Packy McCormick.

Claude Code Is The Inflection Point — SemiAnalysis (newsletter post)

• Type: Newsletter post Not Boring by Packy McCormick
• Author/creator: SemiAnalysis Not Boring by Packy McCormick
• Link/URL:https://newsletter.semianalysis.com/p/claude-code-is-the-inflection-pointNot Boring by Packy McCormick
• Recommended by: Packy McCormick (Not Boring) Not Boring by Packy McCormick
• Key takeaway (as quoted): “4% of GitHub public commits are being authored by Claude Code right now,” with a projection that it could reach “20%+ of all daily commits by the end of 2026” Not Boring by Packy McCormick.
• Why it matters: It’s a specific metric + trajectory claim that can recalibrate how quickly you expect AI-assisted coding to show up in day-to-day software production Not Boring by Packy McCormick.

Tool Shaped Objects — Minutes (essay)

• Type: Article/essay Not Boring by Packy McCormick
• Author/creator: Minutes (publication) Not Boring by Packy McCormick
• Link/URL:https://minutes.substack.com/p/tool-shaped-objectsNot Boring by Packy McCormick
• Recommended by: Packy McCormick (Not Boring) Not Boring by Packy McCormick
• Key takeaway (quote highlighted in Not Boring):

• Why it matters: A compact framing for evaluating tools, dashboards, and workflows that optimize for the appearance of progress rather than actual outcomes Not Boring by Packy McCormick.

Strategy + history: control, institutions, and who makes the call

The Control Revolution — James R. Beniger (book)

• Type: Book Not Boring by Packy McCormick
• Author/creator: James R. Beniger Not Boring by Packy McCormick
• Link/URL:https://www.amazon.com/Control-Revolution-Technological-Economic-Information/dp/0674169867Not Boring by Packy McCormick
• Recommended by: Packy McCormick (Not Boring) Not Boring by Packy McCormick
• Key takeaway (as shared): Packy says he’s only started it, but flags Beniger’s idea that modern information technology emerged as a response to industrial scale and complexity—an industrial-era “crisis of control” where information/communication innovations lagged behind energy and manufacturing advances Not Boring by Packy McCormick.
• Why it matters: It’s a lens for thinking about why information systems and coordination mechanisms proliferate when production and complexity accelerate Not Boring by Packy McCormick.

The Making of the Atomic Bomb — Richard Rhodes (book)

• Type: Book 20VC with Harry Stebbings
• Author/creator: Richard Rhodes 20VC with Harry Stebbings
• Link/URL: Not provided in-source 20VC with Harry Stebbings
• Recommended by: Rory (20VC panel) 20VC with Harry Stebbings
• Key takeaway (as shared): Rory describes a lesson from reading it: military leadership “didn’t give a rat’s ass about the scientists” and that expecting “the luxury of getting to be part of the decision” is “unrealistic” 20VC with Harry Stebbings.
• Why it matters: A grounded reminder about institutional power and decision rights—especially relevant when technical teams assume they’ll control downstream use of what they build 20VC with Harry Stebbings.

Owen (Intercom) on building Fin on top of an existing business (article/blog post)

• Type: Article/blog post (not linked in-source) 20VC with Harry Stebbings
• Author/creator: Owen (Intercom) 20VC with Harry Stebbings
• Link/URL: Not provided in-source 20VC with Harry Stebbings
• Recommended by: Rory (20VC panel) 20VC with Harry Stebbings
• Key takeaway (as shared): Described as a “really great piece” about taking an existing business, “gut[ting] it,” and building Fin on top—emphasizing not just technical work but the fortitude to bet on the new thing, including “probably a year of feeling ridiculous” amid customer pressure 20VC with Harry Stebbings.
• Why it matters: A candid reference point for leaders attempting major AI-driven product transitions inside a live, legacy business 20VC with Harry Stebbings.