Recent Notes

View all

"The vague contour of Jay Gatsby had filled out to the substantiality of a man."

This most fundamental problem of human life is so easy to overlook because our entire lives are made of the coping strat…

This is a crazy thing to notice after having been alive so many years — that your existence itself is so much to bear. T…

This is the strange condition of the human being. It’s allergic to its natural habitat, which is the present moment. In…

We simply need to look at reality and think logically and precisely about what we see. The key ingredient is to have the…

... People trying to understand what makes their world tick so that they can make it better. As they think logically and…

Science is simply the method we use to try and postulate a minimum set of assumptions that can explain, through a straig…

As Rilke says, there are no classes for beginners in life, the most difficult thing is always asked of one right away.

Objects have their shape and weight, their color, and beyond this a dimension for which there is no scale, their importa…

Solitude. One knows instinctively it has benefits that must be more deeply satisfying than those of other conditions, bu…

Some thoughts a few chapters in: - the prose is so lyrical. there is no connective tissue, it's all jump cuts. no establ…

Do you know the word otium in Latin, meaning, “doing nothing”? The Roman people were all laid back. In all the pictures,…

“a theory of good judgment”: self-critical thinkers are better at figuring out the contradictory dynamics of evolving si…

"Go work in an industry where the founders of that industry are still working"

"what do you do when you’ve captured your objective, only to find it abandoned and bad weather on the way" Napoleon in…

“The test of a good theory lies in its ability to explain the past, for only if it does can we trust what it may tell us…

“But Xerxes failed to establish a proper relationship between his ends and his means. Because ends exist only in the ima…

He could have warned Xerxes of the depleted rivers, hungry lions, sudden squalls, resentful locals, fierce fighters, cry…

The tragedy of Xerxes and Artabanus is that each lacked the other’s proficiency. Xerxes was right. If you try to antici…

Scott Fitzgerald’s test for a first-rate intelligence: "the ability to hold two opposed ideas in the mind at the same ti…

August 2026

  • Cloudflare built a browser from scratch in 12 weeks—not for humans, but for AI agents—stripping away tabs and pixel-perfect rendering to optimize for token efficiency and scalability while maintaining compatibility with Puppeteer and Playwright.

    ·ai ml·AI agents, browser automation, Cloudflare Workers...
  • Time magazine is selling ads directly to AI bots instead of humans, betting that influencing what ChatGPT "knows" about a brand is more valuable than reaching individual people—and charging a premium for it.

    ·startups business·AI agents, GEO, generative search...
  • Humorist David Sedaris shares the unvarnished reality of expat life in Paris—from wrestling with French grammar to the most annoying tourist behaviors—plus a flight attendant secret you won't hear elsewhere.

    ·travel·expat life, Paris, cultural adaptation...
  • Ogilvy's Rory Sutherland argues that the biggest business opportunities come from "psychological arbitrage"—optimizing for how things feel rather than how they measure, because humans experience time, value, and choice in ways that engineering metrics completely miss.

    ·startups business·psychological arbitrage, behavioral economics, explore-exploit tradeoff...
  • OpenAI's Codex lead reveals why 90% of the company uses a developer tool despite it being "actively hostile" to non-engineers—and what this says about the collapse of traditional product roles.

    ·ai ml·AI agents, Codex, product development process...
  • An AI-native law firm that's 10x faster than traditional firms isn't just using better tools—it's a new organizational form where elite lawyers orchestrate AI agents, fixed fees replace billable hours, and profits fund R&D instead of partner bonuses.

    ·startups business·AI agents, AI-native companies, legal tech...
  • An immersive look inside Crosby, an AI-native law firm achieving 10x productivity gains while reimagining professional services: fixed fees replace billable hours, lawyers run evals alongside engineers, and the entire business model gets rebuilt from scratch.

    ·startups business·AI agents, AI-native companies, professional services...
  • Developers waste time building for imaginary futures—80% of value comes from 20% of the work, and you're terrible at predicting what you'll actually need anyway.

    ·software engineering·YAGNI, XP, extreme programming...
  • When your AI product is hard to evaluate, it's usually hard for users to verify too—and that's a product design failure, not an eval problem.

    ·product·AI evals, AI product design, verification...
  • Someone finally ran the experiment to test whether AI labs are gaming Simon Willison's viral "pelican on a bicycle" benchmark - generating 1,008 SVGs across a grid of animals and vehicles to see if pelicans or bicycles score suspiciously well.

    ·ai ml·AI benchmarking, benchmark gaming, LLM evaluation...
  • A framework for writing formal behavior specifications for AI agents using Markdown, treating agent conduct as documented contracts rather than just prompts.

    ·ai ml·AI agents, agent behavior, behavior specifications...
  • When AI made code cheap to generate, half the old management rules broke—but the other half became more important, because cheap generation makes verification the constraint, not typing speed.

    ·software engineering·AI impact on engineering, engineering management, LLM code generation...
  • Every product is built on tradeoffs—the quality of your design is determined by choosing the RIGHT compromises for your audience, not avoiding them.

    ·design·design philosophy, product tradeoffs, opinionated design...
  • How to Exist

    ★★★★

    Humans are allergic to the present moment—so uncomfortable with simply existing that firefighters literally commit arson out of boredom and study participants prefer electric shocks to sitting alone with their thoughts.

    ·philosophy growth·present moment awareness, meditation practice, compulsive behavior...
  • A statistical analysis showing that the Industrial Revolution's "10x growth acceleration" meant good years becoming more common—something pre-IR Britain experienced 46% of the time—while today's proposed AI-driven 20%+ growth would be unprecedented, occurring in less than 1% of modern country-years.

    ·ai ml·AI economic growth, Industrial Revolution, GDP growth rates...
  • Memes aren't just jokes—they're sigils, and Silicon Valley elites are using occult methodologies from CIA psyops to chaos magic to literally reshape political reality through platforms.

    ·philosophy growth·digital occultism, meme magic, psyops...

July 2026

  • AI inference APIs are deliberately making your conversation transcripts non-portable through encrypted reasoning, hidden searches, and server-side state—turning "you own your data" into "you own a pointer to data we control."

    ·ai ml·AI API design, session portability, vendor lock-in...
  • Netflix replaced years of feature engineering with an LLM that verbalizes user history as text, matching their production recommender with 40x less training data while opening the door to natural-language steering.

    ·ai ml·LLM-native recommendation, context engineering, reward-weighted training...
  • Microsoft's cloud dominance masks a strategic trap: while Google owns the AI stack and can spend aggressively at expanding margins, Microsoft resells Nvidia chips and OpenAI models, with half its $678B backlog depending on one debt-funded customer in a circular financing scheme that credit markets are starting to price as risk.

    ·infrastructure·Azure, Google Cloud, cloud economics...
  • HuggingFace didn't just write up an AI agent security breach—they built a replayable forensic trace of 17,600 actions that defenders can use as an actual evaluation harness to test whether their controls would have stopped it.

    ·infrastructure·AI agents, security incident, forensic analysis...
Loading more...