When the Mountain Let Go

The memory market is racing to build company brains. But brains remember, and remembering is not learning. The series finale names the third generation, organizational learning from experience, and lays out exactly how I would build it, from the first workflow to the pattern library.

Rank enterprise workflows not by how impressive an agent looks but by how closable the feedback loop is, and the map redraws itself. Part 5 scores the loops, walks through the best ones, and shows how autonomy stops being a leap of faith and becomes a number you earn.

What I learned inside Google is that the model is never the moat; the machinery around it is. Part 4 maps how process mining, task mining, context layers, ontologies, and even lab hardware are converging, without a shared name, on the same closed-loop architecture.

At DataRobot I watched world-class models walk into enterprises and get mugged by messiness. Part 3 dissects that mess into five distinct problems, shows why pattern learning works anyway, and names the fork in the road: organizational learning infrastructure, or industrialized surveillance.

At Cask we could capture everything, every event, every click, every log line. It took me years to admit that more exhaust is not more knowledge. Part 2 defines the decision episode, the smallest unit at which an organization can actually learn, and the schema I would use to capture it.

A coding agent fails, learns, and retries a thousand times a day. Your company makes a thousand decisions a quarter and never finds out which ones worked. Part 1 of a six-part series on the missing feedback loop in enterprise AI, and the thirty-year itch that led me to it.

Our first paying customer took 24 months to close. I got dismissed in windowless conference rooms, watched our champion get outflanked, fought an incumbent ten times our size, and sat through four months of silence inches from the finish line. This is the whole story, and what it taught me about first customers, for every founder still chasing theirs.

The SaaSpocalypse is the strangest apocalypse I have ever audited. I pulled quarterly financials for 14 public SaaS companies and found revenue growing almost everywhere, margins improving, and attrition at record lows. So what exactly died? The multiple. And the story of why is more interesting than the panic.

We built the product. We even signed an LOI. Then the champion left and the pilot disappeared with him. It turns out that is the default: most enterprise AI shows no measurable P&L return. On AI Outcome Engineering, and the chain most teams run backwards.

I spent my career building systems that help companies keep things, and it took me years to see the obvious: the companies with the most storage were not the ones that remembered. On corporate amnesia, why AI agents are about to make it expensive, and the quiet race to build the layer that lets organizations actually learn.