Eugene Kungurov
Finding simple solutions to complex problems, or at least simple language to explain them.
Recent Articles
-
AI Agents Learn From Your Worst Code First
AI agents pattern off the code they read, and verbose generated or legacy code has ten times the surface area of clean code. Encapsulation keeps the agent focused on code that reflects the team's conventions.
-
Building AI Agents: Short Narrative Beats More Instructions
Neither narrative nor constraints work alone when building AI agents. Combine them — context first, then directives — and the agent fills gaps with judgment instead of guessing blind.
-
Role Separation Turns AI Agent Review Into Actual Review
A single agent writing code and reviewing it isn't real review. The fix is isolated context, adversarial intent, and agents scoped to one role each.
-
Anchor AI With Precise Language, Not More Instructions
The real problem isn't AI tooling — it's finding the right place for the right context. A few precise sentences reshape everything downstream.
-
AI Supply Chain Risk: The Missing Category
We talk about AI alignment risk and AI misuse risk. We're missing a third category: what happens when AI becomes a supplier in its own infrastructure supply chain.