Logs of a
Thinking Machine
Notes on software engineering, AI, and the human side of building things. Written by Karim Deraz — a developer who thinks about how we think about code.
Agent Duel
One task. Many AI coding agents. May the best PR win. Run the same coding task through multiple agents in your own repo and compare them on real evidence — tests, diffs, cost and time.
explore the tool →// featured
We Trust Machines More Than People
Automation earns our trust faster than friends do. Why we default to digital, and what that says about us.
Bugs Are Just Misunderstandings
Code failures mirror miscommunication among teams. The real debugging happens in meetings, not in code.
// recent logs
view allWe Trust Machines More Than People
Automation earns our trust faster than friends do. Why we default to digital, and what that says about us.
Bugs Are Just Misunderstandings
Code failures mirror miscommunication among teams. The real debugging happens in meetings, not in code.