In today's article, I write a summary of my experiences working with an LLM coding assistant for a week. Surprisingly, it's less negative than the article I expected to write. Unsurprisingly, it is still very, very negative and comes to the conclusion that LLM agents probably shouldn't be used for anything

Lobsters is a community-driven platform for sharing and discussing links to articles, tutorials, and projects related to technology and programming. Readers can learn about a wide range of topics, from software development and system administration to cybersecurity and artificial intelligence. With user submissions, comments, and voting, Lobsters provides a platform for collaborative learning and knowledge sharing among technology enthusiasts.

Lobsters

A week-long hands-on evaluation of opencode, an LLM-based coding framework, reveals mixed results. While the tool excels at generating boilerplate code and reducing activation energy for solo developers, it struggles significantly with greenfield projects, DevOps tasks, and test generation. The author finds that effective use requires expert-level scaffolding, extensive manual review, and rigorous git discipline. Despite some utility for small personal projects, the tool's verbosity, tendency toward generic code style, ethical concerns around training data, and inability to guarantee code correctness make it unsuitable for production use. The promised productivity revolution remains unrealized.

My week with opencode

Why is a notorious LLM critic doing this?