Elevator Pitch

• After building multiple real apps with Claude Opus 4.5, the author concludes AI coding agents can effectively replace developers for substantial end-to-end work—if you stay vigilant about security.

Key Takeaways

• Opus 4.5 feels qualitatively different from prior agents: it iterates via CLI, reads errors/logs, and fixes issues without extensive handholding.
• The author used Opus 4.5 to rapidly ship multiple apps (Windows utilities and iOS apps) including full backend integrations (Firebase, auth, storage, scheduled posting).
• Code may no longer need to be optimized for human readability; the author advocates “AI-first” code that’s explicit, linear, low-coupling, and easy for LLMs to regenerate—while manually scrutinizing security (API keys, auth, sensitive data).

Most Memorable Quotes

• “Today, I think that AI coding agents can absolutely replace developers. And the reason that I believe this is Claude Opus 4.5.”
• “Why does a human need to read this code at all? I use a custom agent in VS Code that tells Opus to write code for LLMs, not humans.”
• “Just make sure you know where your API keys are.”