A developer shares their two-month journey adopting AI tools for coding, writing, and daily tasks. They describe their setup using OpenRouter and OpenCode, experiencing both successes (one-shot implementations, brainstorming system designs) and frustrations (prompt anxiety, review overhead, nondeterministic behavior). AI proved most valuable when the developer already had a mental model and used it to execute tasks, for checking comments and reviewing changes, and for brainstorming system architecture. Coding tasks required significant hand-holding and chunking into smaller subtasks. The experience highlights practical tradeoffs: cost management with per-token pricing, the superiority of AI for contextual search over traditional engines, and usefulness for low-stakes decisions and language learning.
