I spent $100 in Claude tokens and 1k battles training my AI tank

Name: I spent $100 in Claude tokens and 1k battles training my AI tank
Availability: InStock
Author: mazzystar

by mazzystar·May 13, 2026·5 points·4 comments

Visit Project View on HN

AI Analysis

●●SolidCozyRabbit Hole

Watching the AI fail visibly and fixing its logic is the real hook here.

Strengths

•Replay system lets you audit exactly where the agent's logic failed.
•Feedback loop turns natural language strategy into code updates.

Weaknesses

•Limited to a single game genre; novelty may wear off after a few battles.
•Dependent on LLM quality; complex strategies might confuse the agent.

Post Description

Hi HN, I built AgenTank.

It is a small game where an AI agent writes the logic for your tank. You watch it fight, give strategic feedback, let the agent update the tank code, and send it back into battle.

I have run 1,000+ battles on my own tank and spent about $200 in Claude credits improving it. The part I enjoy most is not just winning, but watching the tank make visible mistakes, thinking of a better strategy, and seeing whether Claude can turn that into better code.