How Do We Stop Starting Over?
Thirty years of watching organizations lose their own knowledge, and why the real question for AI isn't "how do we go faster?"
I've been in enterprise software for over thirty years. In that time I've watched the same movie play out again and again, just with different actors.
In the nineties it was distributed objects. Then service-oriented architecture. Then microservices, containers, serverless. Now AI agents. Each wave brings genuinely brilliant technology, genuinely excited people, and the same predictable outcome: we grab the shiny new thing before we've worked out how the system actually wants to behave.
That pattern, great technology misapplied, is half of what I kept seeing. The other half is harder to talk about, because it doesn't have a catchy name.
It's what happens to the understanding
After every engagement (every modernization, every transformation, every workshop) there's a moment when a team genuinely gets it. They understand how the pieces connect. They understand why decisions were made. They understand what the system is trying to be.
And then it evaporates.
The consultant leaves. The senior engineer retires. The wiki page goes stale. The whiteboard photo gets archived. The reasoning behind the architecture lives in exactly three people's heads, and one of them just left for a competitor.
Six months later, a new team shows up and starts from zero. Again.
Methods help. Until they walk out the door.
Over the years I built methods to fight this: workshop techniques for surfacing how a system wants to be decomposed, templates for capturing what a team had discovered before it disappeared, and eventually a full modernization method that I've taught to teams around the world.
They worked. But they were still manual. Still human-speed. Still dependent on the right people being in the room, and on somebody writing down what they said before they left.
What I really wanted, for most of my career, was for the understanding to compound. Not reset. Not evaporate. Compound.
Every engagement making the next one smarter. Every decision carrying its reasoning forward. Every system understood not just as code, but as intent.
Why now
For the first time, the tools exist to do this properly. AI can read the meeting transcript, the forty-page contract, the legacy code, and the half-finished diagram, and turn them into structured, connected knowledge that stays current and that people and machines can both act on.
That's why I built Lapis. Not because AI is cool (it is), and not because agents are the future (they are). It's because after thirty years of watching organizations lose their own knowledge, I finally had the tools to build something that could hold it, reason over it, and make it grow.
What this means if you run a business
If you're a private equity partner, you've bought companies where the most valuable operational knowledge lived in two people's heads. If you run a mid-sized company, you've paid consultants to learn your business, twice.
The question isn't "how do we go faster?"
It's "how do we stop starting over?"