I've spent 20+ years building AI products at Google, DoorDash, LinkedIn, and a startup I founded in 2012. These are the things I've learned that I wish someone had written down earlier.
Latest
The Algorithm Was Never Listening
Recommendation systems cannot detect authenticity. They run a short, proxy-driven bet on your content and lock in fast. Understanding the bet changes what you do before you publish.
What I Prefer
Marc Andreessen posted his AI custom prompt. Here is mine.
Neither Tool Nor Colleague
One camp believes AI will restructure everything. Another believes the hype will fade. Both resolve cognitive dissonance rather than engage with what the technology actually is.
Finding the Unit
Whether you build a product or a platform is not a strategic choice. It is a consequence of what your unit of value actually is. Finding the unit is the work. Everything else follows.
AI Speaks in Language. It Reasons in Statistics.
AI operates statistically but presents linguistically. Three quantitative tools — base rates, precision and recall, and statistical significance — give you a way to reason about AI outputs that the language layer alone cannot.
Non-Determinism in Enterprise AI: What It Actually Is, Where It Comes From, and What To Do About It
Statistical, probabilistic, and non-deterministic are three distinct properties of AI systems. Conflating them is one of the most common and costly mistakes in enterprise AI adoption.
Who you hire, who you grow, who you promote — and how AI gets this wrong
On noise, bias, and what Amazon's failed hiring tool actually teaches us about deploying AI into people decisions.