// automation

Source: Azeem Azhar, Exponential View

A Ukrainian combat drone pilot's observation that soldiers require complete retraining after eight-month absences shows how fast operational change moves in modern warfare. Drone tactics and countermeasures now evolve faster than traditional military doctrine cycles can absorb, forcing real-time adaptation. This mirrors how software development has compressed hardware refresh timelines. Forces that can institutionalize continuous retraining and tactical iteration gain a structural advantage—organizational agility is becoming a scarce military asset.

Source: Yaschamounk

Garicano's framing sidesteps the complement-or-replacement binary by naming the actual economic mechanisms at play—which Silicon Valley's techno-optimists routinely miss. The gap between venture-backed automation rhetoric and real labor market outcomes isn't a timing problem. It reflects how AI deployment decisions depend on institutional constraints, wage structures, and competitive dynamics that tech founders have little reason to understand. What matters is whether organizations choose to augment workers or eliminate roles. That choice is driven by economics and power, not capability. That distinction determines whose jobs survive.

Source: LessWrong

The cycle of AI hype has accelerated to the point where labs release capabilities (coding agents, multimodal models, reasoning systems) faster than institutions—companies, regulators, educational systems—can integrate or respond to them. Each new capability class triggers speculative frenzy and "new era" declarations before the previous wave has been debugged or deployed at scale, leaving organizations perpetually playing catch-up. The pressure has shifted from AI capabilities to market and institutional absorptive capacity: what are these tools actually for.

Source: Marginal REVOLUTION

A Financial Times survey of 4,000 US and UK workers shows AI tools concentrating among high earners: over 60% of top earners use AI regularly, while adoption rates decline steeply down the income ladder. Higher-wage workers gain productivity multipliers from ChatGPT, Claude, and specialized tools that lower-wage workers lack, automating the routine work that historically opened paths to better jobs. Without deliberate effort to distribute AI literacy and tool access downward, this skill gap will harden into structural wage inequality within 3-5 years.

Source: Financial Times

The adoption gap isn't about access or training. Senior and well-paid workers are pulling ahead because they can afford to experiment with AI tools, have time to learn them, and work in roles where AI augments rather than replaces their labor. This compounds existing advantage: those already positioned at the top of the labor market gain productivity boosts that widen pay and opportunity gaps, while workers in lower-wage roles face displacement without resources to retrain.

Source: Quartz

Meta's experimentation with an AI version of Mark Zuckerberg for internal use exposes a real corporate tension: executives want to scale their decision-making and communication without the friction of actual delegation, but an AI simulacrum of leadership creates a liability black hole when things go wrong. The move reflects anxiety about the present, not vision for the future—a shortcut for companies unwilling to build management depth, train middle layers, or distribute real authority. If decisions made by an AI trained on a CEO's patterns cause harm, who bears responsibility, and what does trust in leadership mean when the leader isn't present?