Showing 9 of 18 posts
Warren Buffett's punch card rule imagines an investor limited to 20 decisions for life. The scarcity forces deep research, concentration in your best ideas and the patience to do nothing, the opposite of what frictionless modern markets encourage. You don't need a literal card. Just invest as if you had one.
AI lowers the cost of specialised execution, so the bottleneck moves from doing to deciding. That favours breadth, but depth still matters wherever the model breaks. The durable profile is both: wide enough to use AI across domains, deep enough to go beyond it. Intelligence is cheap. Judgement is not.
AI is a multiplier, not an equaliser by default. Early gains flow to capital owners, middle-income work hollows out, and power shifts from labour to capital, within countries and between them. The same tools could widen opportunity, but only if access, education and ownership are designed for it.
For emerging markets, AI could lift agriculture, healthcare and education, or gut the outsourcing sectors they depend on and deepen reliance on foreign platforms. The technology doesn't decide which path a country takes. Deliberate investment in talent, infrastructure and institutions does.
Boards hold decades of experience and little proximity to the present, while creators, power users and recent graduates shape perception with no seat at the table. A Community Advisory Council sitting alongside the board, not replacing it, would give leadership a live read on the people it serves.
LLMs are smart, but intelligence is not wisdom. They give wise-sounding answers with no lived experience, no values and no stake in the outcome, which makes them risky to trust with high-stakes calls. The goal isn't machines that replace human judgement. It's systems that sharpen it.
Y Combinator didn't fail. It succeeded so completely that it made itself less necessary. Capital is abundant, networks are internet-native, batches are huge and distribution beats pedigree, so the best founders no longer need it. YC is still valuable, especially for first-time founders. It just isn't the gateway.
Real data is expensive, scarce and regulated, so a growing share of what trains AI is generated rather than collected. Synthetic data eases the bottleneck, the privacy problem and the shortage of rare edge cases, and adds risks of its own: drift, bias and model collapse. Validation becomes the constraint.
Your birthday gives one number, but you carry three ages: calendar, biological and mental. Calendar age counts the years. Biological age tracks how well your body works. Mental age is your curiosity, maturity and adaptability. The gaps say more than the number, and two of them are yours to change.