The second monthly link post. Same idea as last month: the links I saved in June. Heavier on AI than I expected going in, but there is some history, a security writeup, and a couple of older classics worth revisiting.

This month

• When AI builds itself, Jun 2026
  • Anthropic’s own account of AI starting to automate AI research, and why recursive self-improvement is the thing they say they are now watching most closely.
• Writing is Thinking, 16 Jun 2025
  • A Nature editorial on why you should not hand your writing to a model wholesale: the act of writing is where the thinking actually happens, not a chore around it.
• The Kaiser and a “Mediocre Man” Theory of History, 4 Apr 2024
  • A case that Wilhelm II mattered precisely because he was mediocre, and that history sometimes turns on the ordinary failings of whoever happens to hold power.
  • It reminded me of Mark Twain’s “Luck”, though the two stories work in opposite directions. Twain’s fool is carried upward by chance and mistaken for a hero, so the error is in how we read him. The Kaiser is no genius either, yet he really does bend events, so the mediocrity itself is the cause.
• Nobody Ever Gets Credit for Fixing Problems that Never Happened, Repenning & Sterman, Summer 2001
  • The classic systems paper on why organizations get stuck firefighting: fixing root causes is invisible, so the work that prevents crises is never the work that gets rewarded.
• A backdoor in a LinkedIn job offer, 15 Jun 2026
  • A developer walks through a fake recruiter’s “coding task” that would have run a remote backdoor on npm install. A tidy anatomy of a social-engineered supply-chain attack.
• Has AI Already Killed How-To Nonfiction?, 12 Jun 2026
  • Tim Ferriss shows his own catalog’s sales falling off a cliff as models commoditize how-to information, and argues the survivors will sell transformation and community rather than facts.
• 1,000 True Fans, Kevin Kelly, 4 Mar 2008
  • The essay that keeps getting more relevant: you do not need millions, just a thousand people who will buy whatever you make. Read it next to the Ferriss piece above.
• The Wholesale Plagiarism of Obscure Sorrows, Andy Baio, 18 Jun 2026
  • An agency AI-cloned John Koenig’s entire Dictionary of Obscure Sorrows into a slicker, better-ranked site and monetized it. The new shape of plagiarism.
• Why Coding Is Solved, and What Comes Next, Boris Cherny, Anthropic
  • The Claude Code lead argues the bottleneck has moved off writing code and onto everything around it: taste, review, and deciding what to build.
• Coding is no longer the constraint: scaling devex to teams and agents, Spotify
  • How a large org rebuilds its developer platform once both humans and agents are shipping code, and coding stops being the thing that limits you.
• How Lovable vibecodes production software at scale, Lovable
  • A look inside how one of the fastest-growing AI products actually ships, from prompt to production.
• Stanford Leadership Forum 2026: a conversation with Ken Griffin, Ken Griffin at Stanford
  • The Citadel founder on markets, talent, and the long work of building an institution.
• We Need an International Treaty to Ban Superintelligence, Andrea Miotti, 24 Jun 2026
  • A case for a verifiable international ban on building superintelligence, on the nuclear-nonproliferation model, while monitoring the compute is still feasible and the window is still open.
• Half-Baked Product, 2 Jul 2026
  • A parable of an oven startup that bolts on feature after feature to chase every sale while never fixing its core algorithm. On quietly trading the mission for short-term revenue.
• My Antichrist Lecture, Scott Alexander, 22 Oct 2025
  • We were discussing with friends how the world feels like it is accelerating toward something, the way technology did in the run-up to WWII, which got me wondering whether there is any resemblance between AI and the Antichrist. To my surprise, the first result when I searched on Google was a Scott Alexander blog post (which I assume is satirical) on similarities between Anthropic and the Antichrist.