The Rise of Async Programming

Summary

19 August 2025Ankur Goyal I spend a decent amount of time reviewing code I didn't write. An AI agent takes a detailed problem description, writes code (primarily Typescript, Rust, and Python), adds tests, and commits the changes to a branch. I tap back in when everything's ready for review. This used to feel like a futuristic scenario, but it's how I work now, and it's how many developers are starting to work. The shift is subtle but powerful: instead of writing code line by line, we're learning to describe problems clearly and let tools solve them in the background. This version of "async programming" is different from the classic definition. It's about how developers approach building software. The workflow looks like this: Define the problem clearly. Write a detailed specification of what needs to be built, including edge cases, constraints, and success criteria. Hand it off. Delegate the implementation to an AI agent, a teammate, or even your future self with comprehensive notes. Return later. Come back to review results, provide feedback, and decide on next steps. The key difference from traditional programming is the time separation between problem definition and implementation. Instead of immediate feedback loops, you have background problem solving driven by clear requirements and automated verification. Async programming is not vibe coding. Vibe coding enables you to write code without getting into the nitty gritty details. Async programming is a workflow for developers to solve more complex problems simultaneously, while still understanding the details of the code being written. You're still architecting solutions, reviewing implementations, and maintaining a codebase. You're just not typing a vast majority of characters yourself. For async programming to work in practice, you need three things: a clear definition of the problem you're solving, a way to automatically verify your results, and human-driven code review. The quality of your problem statement d...