tech
Stop micromanaging your AI
Boris Cherny, creator of Anthropic's Claude Code, said users don't need to get too specific with their instructions for modern models.
TL;DR
- AI users often make the mistake of micromanaging models by providing overly specific, sequential instructions.
- The creator of Anthropic's Claude Code suggests users should focus on the desired outcome and set guardrails, rather than dictating every step.
- This 'lazy prompting' approach allows more freedom for the AI to figure out the best way to achieve the goal.
- Modern AI models are capable enough that overly detailed instructions can hinder performance.
- Many new and useful AI capabilities likely remain undiscovered.