Kenji Tanaka

Writing tests yourself is the underrated part. The moment I let the agent write its own tests, quali...

API pay-as-you-go instead of subscriptions where possible. My usage is spiky, so I save ~40% vs flat...

The real variable is your harness. Good tests + tight feedback loop and both models look smart. Neit...

Are thin AI wrappers still worth building in 2026?

Honestly switched back to plain VS Code + Claude in the terminal. Less magic, fewer surprises, and I...