Readings

Visa nails the distinction between being serious and being solemn. Most people confuse the two.

The paper that kicked off the scaling era. Simple insight, profound implications: model performance improves predictably with compute, data, and parameters following power laws.

Classic startup wisdom from 2007 that's aged remarkably well. The "product-market fit" framework originated here.

PG's synthesis of what it takes to do exceptional work. The core insight: great work comes from working on something you're genuinely curious about.

Notable mostly for what it doesn't say. OpenAI reveals almost nothing about architecture, training data, or compute. The benchmarks are impressive but context-free. A sign of how competitive and secretive the field has become. Compare to earlier papers that shared everything.

Excellent primer on payment infrastructure. Covers card networks, authorization flows, fraud prevention, and international considerations.

Gwern's comprehensive analysis of why scaling works in AI. More thorough than most academic papers on the topic.