Metrics definitions beat big numbers

xAI’s “600M MAUs” for Grok is a great example of why founders should obsess over metrics definitions, not just big numbers. What they’ve basically done is add up X’s total reach and call it Grok usage. Independent estimates put actual Grok MAUs closer to roughly 30 to 64M. Big difference. One is “people who could theoretically see a Grok button somewhere.” The other is “people who actually use the model.” As founders, we all know this game: - “Registered users” paraded as “active” - “SDK installed” counted as “product adoption” - “Potential reach” quietly labeled as “engagement” It looks good in a deck. Until it doesn’t. If AI labs can mix distribution surface with real model usage, those headline MAUs will end up driving $200B+ valuations, policy debates, and maybe even regulatory thresholds. That is a terrible place to have fuzzy numbers. The solution is pretty boring and very needed: standardized, auditable MAU definitions for AI products. Something like: - Clear event definition of “use” - Time window (e.g. 30 days) - Separation of host-platform users vs model users If we want this ecosystem to stay credible, we need to stop speed-running the adtech-style metrics inflation playbook.