Writing

Field notes from building the verification layer.

Research, incidents, and the decisions behind the Veriva pipeline — written for the engineers, founders, and governance teams thinking hard about how AI-generated code actually ships.

Latest · May 7, 2026 · 10 min read

The Hard Part Is Knowing When the Code Is Done

ProgramBench shows why AI coding needs executable acceptance criteria, not just stronger generators.

Alexander TheruviparambilRead the essay

April 28, 20269 min read
Your AI Code Reviewer Can't Be Your AI Code Generator
How Anthropic's April 23 post-mortem made the case against same-vendor verification.
April 23, 20268 min read
Intuition Without a Second-Guess
Why frontier AI still overreaches, and why the fix can't live inside the model.

Field notes from building the verification layer.

The Hard Part Is Knowing When the Code Is Done

Your AI Code Reviewer Can't Be Your AI Code Generator

Intuition Without a Second-Guess