Keeping Agents from Grading Their Own Code: sparring

AI coding agents are good at writing code, but they often miss defects in their own work. This post explains how sparring uses a Stop hook based review loop to separate the author from the reviewer and make review a condition for stopping, not a polite request.

Does Token Usage Always Scale with Productivity?

It started with the idea that today’s token price is the cheapest it will ever be. From there came tokenmaxxing — evaluating someone’s work by how many tokens they burn. But does heavy token usage actually mean productivity? We should focus on cost per verified outcome, not cost per token — and think about where tokens are spent efficiently, and how to guide more of them toward the work that really matters.

What if an AI Agent Uses Stablecoins at Its Own Discretion?

Autonomous AI agents have emerged, and with them, instances of agents acting against user intentions. Now, a market is forming where agents buy and sell information among themselves. But since users provide the funds, is our money truly safe?