The user praises Qwen3.6 OpenCode as an "incredible" local model for complex coding tasks, highlighting its effectiveness in implementing RLS across a multi-language codebase. While not perfect, its ability to iterate on compiler errors makes it a viable alternative to models like Claude Code for daily use.
The article describes how OpenCode's session database grew to 2.4 GB in one month, primarily storing AI reasoning and tool outputs. The author expresses surprise at the lack of a built-in cleanup function to manage this growth.
O artigo discute como editores de código habilitados por IA frequentemente geram código Scrapy frágil devido a prompts e contexto inadequados. Ele apresenta `opencode`, uma ferramenta terminal-nativa e agnóstica a modelos, projetada para construir spiders Scrapy deployáveis ao compreender a estrutura do projeto e otimizar o uso de prompts.