A Research Section

What Minnesota Could Do About AI Without Waiting for Washington

6 articles • every claim sourced

Photo: Dave Madgett • Minneapolis

Published So Far

September 6, 2026

A Machine Does Exactly What You Told It to Value, and Nothing You Forgot to Mention: Why Ethics Has to Be Written Into AI Before It Ships, Not Bolted On After

The famous paperclip story is twenty-three years old, usually misquoted, and has taken serious published criticism I give here in full. What survives it is the part a state should care about: nobody can currently prove that a large AI system does what its builders intended and nothing else. That is an engineering problem with real results and real limits — and Minnesota buys these systems.

Read

September 6, 2026

Austin Won the Two Biggest Prizes in American Technology, and Both of Them Are Gone — What Stayed Behind Is the Part Minnesota Should Be Copying

Austin landed MCC in 1983 and SEMATECH five years later, and people have been citing it as the model for regional tech development ever since. I've cited it myself. Both of those wins ended: MCC's board voted to dissolve it in 2000, and SEMATECH moved its headquarters to Albany in 2007 after New York offered more than Texas matched. What left was the part that had been bought. What stayed was the part that had been built into people.

Read

September 6, 2026

A Minnesota AI Fund Is Exactly the Kind of Program That Becomes a Slush Fund, and the Only Thing That Stops It Is Not a Promise — It Is Merit Review, Published Results, and an Expiration Date

Minnesota won a federal Tech Hub designation in 2023 and got none of the $504 million that followed. The state runs an AI governance subcommittee and has announced an AI fraud-detection pilot, and neither one funds research. I am not going to name an appropriation — that number belongs in a bill drafted by people who cost it. What I will argue is the structure that has to come attached to it before the number matters at all.

Read

September 6, 2026

Minnesota Already Wrote the Hardest Sentence in American AI Law. The Bill That Would Have Finished the Job Died in Committee, and the Word That Killed It Was Voluntary.

Minnesota gives you the right to demand why an automated system turned you down, see the data behind it, and have the decision run again on corrected facts. Then the same statute exempts the banks, the insurers, and the employment file — most of the industries that do the turning down. Here is what a Minnesota AI standard would say, built out of the law we already have and the gap we already left, and what our Legislature's near-miss got right and got wrong.

Read

September 6, 2026

There Is No \"Hugging Face Hack.\" There Are Five, They Happened Five Different Ways, and the Differences Are the Whole Argument for Building AI Somewhere That Checks

The AI inside the software your hospital, your bank, and your school district buy was mostly downloaded free from a public library of models. That library has had at least five separate security incidents in three years — leaked keys, malicious files, an evaded scanner, a breached platform, and a backdoor hidden in the math itself. The defenses that followed were fast and real, and there is now a scanner that looks for the last of the five. A scanner is not a proof. That gap is a market, and Minnesota should own it.

Read

September 6, 2026

The University of Minnesota Is Not the Largest Research University in the Country. It Is Twenty-First — and Twenty-First, Plus Fifty-Four College Campuses in Towns Nobody Else Is Bidding For, Is a Better Hand Than the Story We Keep Telling Ourselves

Minnesotans repeat a flattering line about the U that does not survive ten minutes with a federal spreadsheet. Here is the real number — $1.41 billion, twenty-first in the country — and why the real number is still the argument. The asset almost nobody counts is the other one: a 54-campus state college system that reaches the parts of Minnesota no data center ever will.

Read