Long-form, obsessively researched.
Essays, tutorials, and frameworks at the intersection of AI tools and systems thinking.
The Expanding Attack Surface: Security Risks When AI Agents Run Your Business
As AI agents gain autonomous tool access — browsing, writing, executing code — the attack surface expands dramatically. A rigorous breakdown of prompt injection, data exfiltration, and agentic risk for individuals and organizations.
AI in 2026: A Clear-Eyed Map of Real Opportunities and Real Risks
A rigorous, evidence-grounded analysis of where AI creates genuine leverage and where it introduces real risk in 2026 — without the hype or the alarm. Decision frameworks for engineers, founders, and knowledge workers.
The 19% Problem: Why Your AI-Assisted Code Is Slower Than You Think
METR's RCT found developers 19% slower with AI despite predicting 20% faster. A rigorous look at the 39-point perception gap and what it means for your workflow.
The 5.5% Inconvenient Truth About Enterprise AI ROI
McKinsey found 88% of orgs use AI but only 5.5% achieve real business impact. A structural analysis of why most enterprise AI investment produces no measurable value.
The Deskilling Ledger: What You Lose Every Time You Let AI Think For You
CHI 2025 found 40% of AI-assisted tasks involved zero critical thinking. A framework for understanding cognitive offloading costs — and when the tradeoff is worth it.
When AI Becomes a Crutch: The Mental Health Data No One Wants to Talk About
1.2M users/week discuss suicide with ChatGPT. MIT/OpenAI RCT shows AI use correlates with loneliness. A careful look at the evidence — without alarmism or dismissal.
Hinton vs LeCun: How to Think When Experts Completely Disagree
Hinton gives 10-20% odds AI destroys humanity. LeCun calls it complete B.S. A framework for forming calibrated views when credentialed experts reach opposite conclusions.