Bad Analogies

Essay arguing that lazy analogies — especially “they’re losing money like Amazon did” — are dangerous shortcuts that obscure whether a business actually has structural advantages.

Core Argument

McCormick dissects why Amazon’s money-losing years are not a universal template for justifying burn. Bezos had a specific negative-working-capital flywheel, category-by-category scale strategy, and emerged with no real strategic competition. WeWork borrowed the analogy and went bankrupt. Uber’s burn actually did resemble Amazon’s network-effects logic and survived. The AI labs (OpenAI, Anthropic) are burning heavily with $19B+ run rates and improving margins, but they face intense direct competition with similar strategies — unlike Amazon or Uber, no single lab has a clear path to strategic solitude. McCormick questions whether recursive self-improvement creates winner-take-all dynamics or just keeps everyone on the same treadmill. He pushes back on “straight lines on graphs” ASI arguments, predicting excellent software rather than God.

RDCO Relevance

• AI agent economics: the API token margin analysis (50-65% gross on Sonnet, 35-50% on Opus) is directly relevant to our cost modeling for always-on agents
• Moat/trust framework: reinforces that competitive strategy matters more than raw spend — applicable to RDCO’s own positioning against commoditized AI services
• Content strategy: McCormick’s self-aware confession about his own exponential-curve essays is a CopyThat pattern worth noting — vulnerability builds credibility