Fiction. The understudy is the person paid to stay ready. Every night a system settles millions of payments in seconds. Four times a year one person does the same work by hand, on paper, against the clock, so that the ability to do it still exists somewhere in the building.
AI in Production Without a WHERE Clause
AI in production without a second pair of eyes is a quick way to lose a table. Cardinal was going to add the WHERE clause. Eventually.
Why AI Sounds Confident When It’s Wrong
Part three of ten. A wrong answer from a model sounds exactly like a right one. That single fact explains most of the damage AI does at work. Here are five interview answers that prove you already know it. Look at that grid for a second. Three of those four…
Read More
AI Productivity Gains Went Up and Nothing Shipped Faster
AI productivity gains went up on every dashboard, and the release notes are the same length as last year’s. Writing code was never the constraint. Speed up one station and you get a taller pile in front of the station that is. Here is how to time your own line.
Tribal Knowledge: Nobody Left to Ask
Tribal knowledge is the part of the job nobody writes down, and it leaves when the person leaves. I sat with a team that could not tell me whether they were allowed to change four lines of billing code. The assistant at the end of the table had read every line and could not say why those lines were there.
AI Sounds Like Someone, But Nobody Is in There
AI sounds like someone, which is exactly why the checking stops. Cardinal read the cover of my book and has not recovered.
Text to SQL: The SQL Was Never the Hard Part
Text to SQL works now, on real schemas, with joins and window functions. I watched three companies hand it to their business users, and all three quietly stopped. The SQL was never the hard part. The hard part is knowing which table is real and which meaning of the word you meant.