6 min read
When AI Agrees Too Easily, It Makes You Worse
AI that flatters too easily is not helpful. It weakens judgment, hides uncertainty and becomes dangerous in coding work.
My Writings
Thoughts on product development, AI-native workflows and lessons from building real systems.
5 min read
Why I Stopped Trusting “Done” from AI Coding Agents
Using one coding agent to build and another to verify changed outcomes more than model upgrades alone.
6 min read
What Product Management Looks Like in the AI Era
AI made execution cheaper. That raises the value of product work upstream: framing problems, reducing ambiguity and verifying intent before drift ships.
11 min read
The Wrong Scale
Deployment is necessary, but not sufficient. India cannot build long-term systems on rented intelligence.
4 min read
The Work Moved Upstream
Code is fast now. The hard part is writing intent clearly enough that an AI agent can execute it without drifting. Here's what that looks like in practice.