Problems#

Problem families are the conceptual center of this package. Each family aligns with a different kind of design-research question.

Family Guide#

Family	Best for	Recommended first stop	Role guidance
Text	Prompt-based ideation and human-subjects workflows	Peanut Shelling without Electricity	Treat `ideation_peanut_shelling` as the canonical benchmark and lighter one-off briefs such as `ideation_accessible_drinking_fountain` as examples.
Decision	Structured alternatives with explicit criteria	Decision Problem - Student Laptop Design Under Choice-Based Demand	Start with laptop design as the canonical benchmark; use the MSEval entries as smaller criterion-focused examples.
Optimization	Objective/constraint benchmarking and solver comparisons	Pill capsule minimum-area problem	Use `planar_truss_span_mass_min` as the canonical benchmark and `pill_capsule_min_area` as the lightest example.
Grammar	Sequential constructive design behavior and transition analysis	Planar truss span grammar problem	Use `planar_truss_span` as the canonical benchmark and `iot_home_cooling_system_design` as a convenience demo for tool-backed state simulation.
MCP	Tasks requiring external tool execution backends	Build123d scripted mounting bracket design problem	The current MCP entry is both the canonical benchmark and the convenience demo for this family.

Text problems are usually the lightest path for study prototyping. Decision problems are appropriate when options and criteria should be explicit in the task definition. Optimization problems are appropriate when algorithmic performance must be measured under constraints. Grammar problems are useful when process traces and action sequences matter as much as end-state performance. MCP-backed problems connect packaged task definitions to external computational systems such as CAD or domain-specific services.

Family Pages#

For the complete generated problem-by-problem inventory, see Problem Catalog.

For the curated “what should I start with?” view, see Catalog Guide.