// user behavior patterns

Source: Daring Fireball

David Smith's Widgetsmith analytics show the iPhone 15 remains the most-used model among his app's audience, a year after launch. The signal is straightforward: power users and early adopters—the segment most likely to upgrade—are holding onto older hardware. Device longevity is increasing, and incremental improvements aren't justifying $1,000+ replacements. Apple's annual upgrade cycle is slowing for engaged consumers.

Source: Lewis C. Lin’s Newsletter

Source: Pewresearch

Polymarket's narrative as a serious prediction market crashes against the reality that nearly two-thirds of its user base are hobbyists placing micro-stakes. The median user is barely engaged enough to move markets or generate meaningful signal. This inverts the startup's core pitch: a platform built on the premise that decentralized betting surfaces truth has instead attracted the same low-friction, low-commitment audience as any consumer gambling app. Prediction markets may not be a scaling business model separate from entertainment.

Source: Search Engine Journal

Anthropic's Claude is capturing disproportionate growth momentum in AI referral traffic—a 4x increase outpacing competitors like ChatGPT—but the absolute numbers remain marginal compared to established search engines, suggesting Claude is winning within a still-niche ecosystem rather than threatening Google's dominance. AI platforms are now meaningful traffic sources, which means websites must compete for visibility across multiple AI surfaces simultaneously and reckon with discovery models beyond traditional SEO.