I’ve been doing IT for forty years.

I’ve written PL/1, COBOL, RPG, Java, PHP, and now JavaScript, on S/36, DEC, AS/400, PCs, and Linux. I’ve worked in CAD, manufacturing, and primarily in distribution and transportation.

I’ve enjoyed creating software, but have always felt that many things are harder than they should be.

I remember a manager early in my career sharing his estimation rule: double the number, then upgrade the unit. Three weeks becomes six months. Two days becomes four weeks. I thought he was cynical. Decades later, having seen plenty of projects through, I have to admit it’s not the worst heuristic out there.

That’s why I’m excited about the current world. Many of the tasks that used to require days or weeks of detailed work have, with the help of AI, become trivial.

Mocking up a database, hooking up a table to screens, integrating with an identity provider, creating a UI that looks plenty good enough for business apps — all these become a matter of a few conversations with an AI assistant. And this introduces challenges. Does a human need to review each line of code? If changes can be made this quickly, how will they be tested? Will this turbocharged ability to change functionality lead to excessive feature churn? The next decade or so, corporate IT will be working through these.

I’ve been building real systems with AI assistance for the last year or so. In these posts I will share my experiences and my thoughts. I hope you find them interesting and useful. More to come.