Better future

The Fourth R

The Fourth R

Reading, writing, arithmetic: the crisis coverage never leaves the three Rs. But the list is a reduction. The old core of schooling put logic at its gateway, and the slogan dropped it. Reasoning has no score, so it can have no crisis, so it is never deliberately taught. And AI just made the other three nearly free.

The Number That Proved They Were Fine

the number that proved they were fine

Michigan is one of the best universities in the country, and it answered a student distress problem by covering first-semester grades. The number its administrators were looking at is true, excellent, and aggregated at exactly the level where the actual problem disappears. This is how capable people miss a root cause, and it is the same error the college-ROI research made for twenty years.

Who Becomes the Next Partner?

Who becomes the next partner?

Juniors learned the trade by doing the grunt work next to seniors. AI is cutting the grunt work and remote work is cutting the proximity, so the apprenticeship that quietly made partners is breaking, with nothing designed to replace it. Who becomes the next partner?

Nobody Can Tell If You Are Right

Nobody can tell if you are right

Michigan covered its grades and the internet answered in a day. Some replies were careful, most were not, and nobody could tell them apart. That is a market failure with a Nobel Prize behind it. It explains why good arguments go unread, why I chose to argue without standing anyway, and what anyone with an audience can do about it in about a minute.

The Measure They Kept

The measure they kept

Michigan will stop recording first-semester grades to help students explore and to curb a mental health crisis. The pressure is real. The evidence says grading is not what made it, and the university kept the measure for itself.

Fooled by Fluency

Fooled by fluency

The Wall Street Journal spent fourteen years testing stock pickers against darts, and fifty years of fund scoreboards say the professionals rarely beat the plain index after costs. The same belief is reassembling around AI: predictive language models read as insight engines, with no scoreboard at all. Same error, new tape. Here is the test that would settle it, and the reason its result will not.