The Erdős unit distance problem asks how many pairs of points can sit exactly one unit apart. Paul Erdős guessed the answer in 1946, and the guess stood for eighty years. In May 2026, an OpenAI model found a counterexample. I checked the grid trick behind it in SQL Server, spatial queries included.
Outdated Statistics in SQL Server: The Optimizer Is Planning for Last Year
Outdated statistics in SQL Server can slow a query even when its code has not changed. On SQL Server 2025 with the legacy estimator, a stale statistic sent one join to 120,043 logical reads, and a targeted update brought it down to 18,220. Auto update helps, but it waits for a threshold, it samples, and it knows nothing about the rows that arrived today.
The Longer Query: A Sunday SQL Illusion
The longer query seems obvious in this Sunday illusion. Compare two bars in a server aisle, then check your answer.
AI Generated ETL Worked on Day One. Then Came Monday.
AI generated ETL ran clean on the first try, and it took six weeks to find out what it was missing. A late Monday file, a file sent twice, a file full of zeros and mangled names all got through. The ugly old job had a rule for every one of them, and nobody had written those rules down. Now I ask AI for the failure list before I ask for any code.
100 AI Interview Questions and Answers for Data Professionals
My new book is out. It holds the 100 AI interview questions data interviewers ask now, with answers short enough to say out loud. Six of those questions are in this post, free.
Install AdventureWorks and WideWorldImporters: Updated 2026
Install AdventureWorks and WideWorldImporters once, the right way. Which file to download for your SQL Server, where to put it so the service account can read it, the RESTORE FILELISTONLY step, restore scripts for T-SQL, SSMS, Docker and Linux, and the five errors people email me about. Every command was run on SQL Server 2025.
NOLOCK Double Counting: One Employee, Two Lunches
NOLOCK double counting is how three people in a lunch line turn into four lunches. Nothing was inserted. One row moved ahead of the scan and got read twice. I tested it on SQL Server 2025 and share a query that catches it on your own table.
The Migration of a Row: Tagged, Followed, Found Again
I followed the migration of a row through SQL Server 2025, from the first write to the grave. The log gets there before the table. The delete does not erase anything. And when the table is empty and the ghost is gone, the statistics histogram still names the row in full.
Navier-Stokes Problem: What It Is and What AI Found
The Navier-Stokes problem in plain words: what the famous fluid question asks, what AI says it found, and why it is not official yet.