Generating and solving difficult logic puzzles with LLMs

[frogger8]

make a logic puzzle for restaurant staffing. Have 23 employees and create a situation that is insane level hard to schedule the shifts. "I" will try to figure out an answer and verify with you if "I" am correct

The restaurant has 23 employees. The goal is to create a schedule that satisfies all the constraints below. Each shift must be completely staffed, and no rules can be violated.

Employees: There are 23 employees with the following specialties:

Chefs (C): Alice, Bob, Charlie, David, Evelyn Servers (S): Fiona, George, Hannah, Ian, Julia, Kevin, Lily, Mason, Natalie, Olivia Hosts (H): Peter, Quinn, Rachel, Sam Dishwashers (D): Tim, Ursula, Victor

Full details https://chatgpt.com/share/67772c03-977c-8007-acb0-6881fdabae42

Answer generated by claude https://claude.site/artifacts/ce387d7f-cd9a-4440-9ba1-b4cc366a70bd

[talldayo]

This is not what HN is used for. Don't create submissions just to report an AI response.

[frogger8]

Not my intention. Look at my history, haven't posted in 1.5 years. Being able to solve unique problems or not is an important data point for some readers like me.

[brudgers]

Maybe more consistent with HN culture to write something up and link it?

[brudgers]

At first blush, this might be Job Shop Scheduling {1}. But I don't know {2}. If so, it is NP-Hard which means verifying an optimal solution is also in NP.

An LLM might be able to solve a toy scale problem with only four operations and lots of "machines." But I am skeptical it can solve problems of "interesting size" more effectively than numerical techniques because verification is also in NP. {3} Good luck.

{1} https://en.wikipedia.org/wiki/Job_shop_scheduling

{2} "All shifts covered" seems like a requirement for optimality.

{3} Assuming the problem is Job Shop Scheduling.

[frogger8]

Note deepseek FAILED to figure it out