Effect > efficiency

Winkletter • 18 Nov 2026 •
Is AI making us more efficient? It depends on how you measure efficiency. If you measure efficiency in terms of utilization, then people are probably getting a lot of work done. Their tokens are maxxed, and they are shoving a lot of material down the pike.
But if you measure efficiency in terms of throughput, starting with an order coming in and ending with a sale being made, that same AI output is actually inefficient overproduction. AI accelerates problems inherent in how we value work.
- What’s easy to count? Tokens and units produced.
- What really counts? Increasing sales while decreasing cost and excess inventory.
I use AI every day to think about the world and research questions that I have. But all I’m doing is generating excess inventory. Writing is my bottleneck. I don’t write enough, and I keep pouring energy into research, so I am the chokepoint.
Tokenmaxxing. Usage FOMO. Context stuffing. Some people are being measured – or measuring themselves – by how much AI they consume.
It’s Wednesday and I’ve only used 35% of my weekly usage!
Resolve a bottleneck and it moves somewhere else. If AI accelerates production, the bottleneck moves to verifying what the AI produces. AI writes your email for you, but now your email inbox is full of AI-generated messages.
As the businessman Eliyahu Goldratt might say, local efficiency is not global effectiveness. The trick, going forward, will be using AI effectively over the entire workflow.