John Hawley and I started Stale Chips in 2021. Five years since, across the Navy, the Air Force, the Army and the NYPD.
Resource prioritization for NAVFAC. A long-range plan an Army installation is still running on through 2030. Offline, local AI for the places where the network is not coming, ever, no matter how many slides say otherwise.
Everybody says government is slow. Not what I've found. The people in the room are sharp and they want to move. What they've been handed is a fuzzy problem.
Fuzzy meaning nobody can see it clearly. It gets declared at the top, built out of assumptions. Then the people who have to build the solution build it out of their assumptions about which assumptions the top assumed. A lovely spiral of guess and check. A thousand monkeys typing Shakespeare comes to mind.
My favorite example, and I promise this one is real. A guy needs a computer for himself and his colleague, so he requests two. His boss says: too small, not worth my time. He requests five. Too small, not worth my time. He requests . Instantly approved.
A team of two, eighty-three spare computers.
Nobody in that chain did anything wrong.
Every incentive worked exactly the way it was designed to. The problem was just never sharp enough for anybody to see what it was doing.
The other direction, from a corporate room rather than a government one: roughly , in two days. We didn't do the cutting. We put the criteria in front of the people who'd have to live with the answer and they cut it themselves, fast, because for the first time the question was answerable.
Sharpen the problem until the decision makes itself.
Almost nobody is stuck on the answer. They're stuck on a question so fuzzy that guessing is the only move left.