LLMs Don't Reason, They Remember
LLMs don't reason, they remember what reasoning looked like. That's powerful within known domains, brittle outside them. Knowing where that boundary lies—and building systems that extend it—is the real skill now.
The Intelligence Shift
Weekly insights on intelligence-native transformation. Patterns, experiments, and frameworks for building when intelligence becomes abundant.
Latest Articles
View all posts » »Exploring what happens when the constraint that shaped all human organization dissolves.
Are We in an AI Bubble? (And Why It Doesn't Matter)
Smart people disagree violently on whether we're in a bubble. For most business leaders, the answer is simple: it doesn't matter. The capability is already here. What remains is tooling, change management—and knowing where to focus.
The Stack: My Approach to Intelligence-Native Architecture
A companion to The Intelligence Bottleneck Is Ending manifesto. How I'm building the operating system for intelligence-native companies—three layers that transform how businesses create value when one person can orchestrate the work of hundreds.
95% of AI Projects Fail — MIT's Report (And What It Means for Your Business)
Despite $30–40B poured into GenAI, 95% of organizations report no measurable P&L impact. The bottleneck isn't the model — it's the missing process layer.
Your AI Strategy is Missing the Process Layer
METR shows frontier AI can handle ~1-hour tasks at 50% reliability—even with 8 attempts. Everyone's waiting for better models. They're solving the wrong problem. A mediocre model in a great process beats a great model in no process.
How Long Can an AI Work Before Failing?
Start measuring AI by how long it can work before it reliably breaks. METR's new 'time-horizon' study gives leaders a clean way to size capability, reliability, and risk.
The Fast Fashion of Software Has Arrived — But Fast Tailoring Is Coming
SMEs can move faster than enterprise by using Gen‑AI to assemble pop‑up apps and fast‑tailored back‑office tools, pricing on verified outcomes and laying the wiring for agent interfaces.
After R1: The Intelligence Moat Dissolves
DeepSeek's R1 has put high-end reasoning into the open. That single move shrank the visible gap between the frontier labs and everyone else—and it matters for how you buy and deploy AI.
The Intelligence Shift
Don't Get Disrupted
Join 1,000+ leaders getting weekly insights on AI transformation