In this workshop, you're going to prove that better models or spending more tokens isn't the most effective means to making software better. You'll be building a bot that will try and sink my battleships by only prompting your agent to build a bot. Your bot will simultaneously battle 15 other specially designed bots, in a challenge to see who can score the most. You'll be looking to be efficient, as the aim of the game is scoring the most points in a single round, without losing points for being sunk or hit. This requires a lot more than generic strategies, oneshotting or brute force.

If you think that "Throwing fable at it" or "I have the best battleships strategy" are the ways to win, think again. This workshop requires a lot more.

We'll show you how the techniques you learn in the workshop can applied to you day-to-day development processes with agents, you'll come away with:
* Knowledge that Better Models are not better than Better telemetry
* Confidence around vibe coding, when and how to de-risk it
* Experience using an MCP during coding, not just when debugging production issues.
* ... and prizes
* ... and bragging rights

You'll need:
* A coding runtime (the language doesn't matter, Javascript and python should be easy to install, but .NET, Java, Go will all work just fine)
* Access to Honeycomb (signup for a free team, that works just as well