LLM Drill Designer

Idea

Users submit a set of natural-language instructions describing what they want to practice, and an LLM parses the input to generate drill definitions, kit lists, and setup instructions. A conversational interface for building drills without needing to understand the drill builder UI.

Motivation

• LE contact noted that drills are typically distributed as PDFs or plain text — this is how instructors already think about and share drill definitions
• Dramatically lowers the barrier to creating custom drills by meeting users where they are (paste in the PDF/text you already have)
• Instructors and organizers often think in terms of training goals, not data fields — “I want a close-range failure drill with two targets at 5 and 7 yards” is more natural than filling out a form
• Could generate the full package: drill definition, kit, setup instructions, and scoring profile suggestion
• Differentiating feature — no competitor does this

Open Questions

• Server-side LLM call or client-side? (Server-side keeps API keys private, aligns with architecture)
• Which model and how is it called? Direct API, or a dedicated service endpoint?
• How much can the LLM infer vs. what does it need to ask follow-up questions about?
• How does the user review and edit the generated drill before saving?
• Should this work iteratively — “now add a reload” — or is it one-shot?
• Cost implications — is this a paid-tier feature?
• What guardrails prevent generating unsafe or nonsensical drills?