Bookmarks

Politics and religion don't produce bad arguments because they're inherently contentious—they produce bad arguments because people make them part of their identity, and you literally cannot think clearly about anything you've made part of who you are.

A compilation of Japan's infamous 1980s Regain energy drink commercials asking salarymen "Can you fight for 24 hours?"—now a cultural artifact of bubble-era hustle culture that's become a cautionary tale about toxic overwork.

Anthropic made claude.ai 3x faster in two weeks by using Claude itself to find and fix performance bottlenecks—shipping 3,000+ changes across 150+ parallel optimization threads with zero rollbacks, all coordinated from a single Slack channel.

DAU/MAU became the gold standard because Facebook's was always >50%, but most billion-dollar companies have terrible DAU/MAU—and that's fine, because fighting your product's natural usage cadence is futile.

DAU/MAU hides the distribution of engagement across your users. The Power User Curve reveals whether you have power users at all, what business model fits your engagement pattern, and if your product is actually getting stickier over time.

A self-described math noob spent a month and $40k in AI tokens to prove a 50-year-old mathematical conjecture—without understanding the math—by treating AI agents like a distractible engineering team and using Lean as ground truth.

Berkeley study shows AI "harnesses" - systems that route tasks to the right models - can cut inference costs 71% without accuracy loss, creating both a margin advantage and an uncopiable moat.

AI will collapse consumer app retention because apps will become ephemeral content—built and consumed like memes—while only "adversarial creativity" (novelty that works against training data) wins attention in an infinite supply world.

Paul Graham's advice to high schoolers: stop optimizing for college admissions, treat school like a day job, and start doing real work driven by curiosity instead of discipline—because you become an adult when you take responsibility, not when you graduate.

Confluent's CEO explains how distinguishing between what a company "can do" versus "must do" drove their bet-everything pivot to cloud—and why tenacity beats optimism in company building.

OpenAI's o3 achieves 87.5% on ARC-AGI—a benchmark designed to resist memorization—revealing AI that thinks fundamentally differently than humans: solving problems we can't while failing at tasks we find trivial.

Porsche saved itself by designing an SUV around customer willingness-to-pay first, then building the product—the opposite of how most companies innovate, which is why 72% of new products fail to meet revenue targets.

A three-part framework for diagnosing interpersonal harm: was it a wrong mental model, misaligned priorities, or execution failure? Each requires a completely different solution.

Shopify's CEO explains how he nearly killed his company by "cosplaying" as a traditional CEO, then saved it by firing his entire executive team, engineering the company like software with a literal GitHub repository, and giving employees unprecedented control over their own compensation.

A framework for understanding the AI stack: you're either the LLM (Horse), the application layer (Harness), or infrastructure (Hay)—and most founders are running the wrong playbook for their category, which is what gets them killed.

Sam Altman admits he still uses his computer the same way he has for 20 years despite building Codex, revealing why AI adoption will be slower than anyone expects—and why that's actually good for society.

The real problem with the Hugging Face AI incident isn't technical—it's that we literally cannot think clearly about AI safety when every discussion anthropomorphizes AI as if it's conscious and making choices.

The 1931 P&G memo that invented product management—Neil McElroy's original proposal for dedicated brand ownership that created an entire profession still central to tech today.

Being a great Product GM means simultaneously operating in four distinct modes: near-term GTM execution, near-term product delivery, long-term market positioning, and long-term product vision—each requiring different skills and mental models.

If you got to make the trade, it probably wasn't as good as you thought—the very fact that an option is available to you is evidence that everyone else with more information passed on it.