Treat the AI like it's yourself (ryanglover.net)

Treat the AI like it’s you: that’s the practical prompt-engineering tip a developer shared after using an LLM to build a database for their JavaScript framework, Joystick. Instead of terse commands, write prompts as if you were writing specs or onboarding notes for a junior you—explicit, verbose, and complete about assumptions, tools, and expected outputs. Framing the model as a cloned persona makes it easier to transfer your domain knowledge, anticipate gaps, and iterate; the author reports that roughly 99% of the project was generated by the AI using this workflow. Technically, this approach foregrounds prompt engineering, context management, and iterative steering as core development skills: include concrete tool constraints and step-by-step instructions, break work into fresh chats when the model shows “brain fog,” and avoid relying on huge context windows to magically solve everything. The implications for AI/ML practitioners are practical and cultural—LLMs can massively accelerate implementation and prototyping, but require human-structured context, chunking, and verification. Expect a learning curve (new workflows, debug-by-prompting), but once mastered this “clone” pattern can transform productivity for many coding and design tasks.