Wiki

Concepts, frameworks, and research behind the book.

• Agentic Systems Infrastructure Production-grade AI agents require new infrastructure primitives — identity, orchestration, evaluation, and fault tolerance — that don't exist in traditional software stacks.
• AI Agents and Job Redefinition AI agents do not eliminate jobs — they reshape job requirements dynamically, creating new task categories while making static employment predictions unreliable.
• AI as Organizational Force Multiplier AI agents give small, adaptive teams leverage that previously required large organizations — creating a structural advantage for startups and AI-native companies.
• Assessment Innovation Generative AI breaks traditional assessments by severing the link between candidates and their output — skills-based and gaze-based methods are the way forward.
• Attention Decline in Knowledge Workers Sustained focus time in knowledge workers dropped from 2.5 minutes (2004) to 47 seconds (2020), with compounding costs from attention residue.
• Bounty and Spread Brynjolfsson & McAfee's twin forces of the digital age: growing total abundance (bounty) alongside growing inequality (spread).
• Centaur and Cyborg Work Two models of human-AI collaboration: Centaurs divide labor strategically (human does X, AI does Y), Cyborgs intertwine efforts sentence-by-sentence — both require understanding the jagged frontier.
• Cognitive Surrender The progressive impairment of critical thinking caused by habitual over-reliance on AI systems.
• Context Engineering Context engineering is the discipline of structuring information, instructions, and tools so AI agents can perform domain-specific work reliably.
• Enabling vs Replacing Technologies Technologies either complement human labor (enabling) or substitute for it (replacing) — this distinction determines social acceptance and political resistance.
• Enterprise AI Adoption Lag 95% of enterprise GenAI pilots fail — not because the technology doesn't work, but because organizations are structurally unable to integrate it into ossified processes.
• Expertise Democratization AI can invert the computerization trend by extending expert-level decision-making to workers with foundational training — rebuilding the hollowed-out middle class.
• Eye-Tracking and Decision Making Gaze patterns reveal unconscious cognitive processes during decisions and cannot be faked or coached.
• Fictitious Commodities Polanyi's argument that labor, land, and money are not true commodities — treating them as such destroys social fabric.
• Judgement vs Knowledge in the AI Era AI reduces the value of what we know and increases the importance of how we think — fewer tasks, more judgement, higher stakes.
• Shadow AI and Organizational Enablement Workers adopt AI faster than organizations can sanction it — banning AI creates shadow usage and data risks, while radical enablement channels the inevitable adoption safely.
• Software Design for the Agent Era Software design is shifting from user-centric to dual-audience (agents + humans) — APIs become primary interfaces, agent infrastructure becomes foundational, and the UX paradigm inverts.
• Technological Revolution Cycles Perez's framework: every major technology follows a 50-60 year cycle of irruption → frenzy → turning point → synergy → maturity, with AI now entering frenzy.
• The Direction of Technology Acemoglu's framework: technology's impact on prosperity depends not on capability but on whether it's directed toward replacing workers or augmenting them — a political choice, not a technological one.
• The Double Movement Polanyi's thesis that market expansion always triggers society's self-protective counter-movement, creating a fundamental tension in capitalist systems.
• The Expanding Work Frontier AI agents will primarily execute tasks humans never did before — the real ROI is in expanding what organizations can do, not just making existing work cheaper.
• The Jagged Frontier AI capability has a jagged, unpredictable boundary — spectacular inside the frontier, catastrophically wrong outside it — and workers who can't map this boundary make worse decisions than those without AI.
• The Productivity Paradox Technology produces productivity gains only after organizational restructuring catches up — electricity took 40 years, computers took 25, AI may follow the same pattern.
• The Technology Trap (Concept) When those displaced by technology have political power to block it, progress stalls — a recurring pattern from guilds to modern automation anxiety.
• The Two Machine Ages The steam engine augmented physical power (first machine age); computers and AI augment mental power (second machine age) — both bend the curve of human history.
• Vibe Coding and the New Software Labor AI coding tools are reshaping who builds software and how — but productivity gains are unevenly distributed, and the technical debt question remains open.