Introducing Fullstack-Bench

Coding agents can do more autonomously when they write code that has tight, automatic feedback loops; use systems that express everything in standard, procedural code; and have access to strong, foolproof abstractions.