That's a fantastic strategic question about building a truly future-proof lights-out factory! Based on current automation trends and lessons from existing implementations, here's my take on what to overspec and underspec:

I'd definitely overspec these by 50%:

1. Power Infrastructure & Connectivity - Running 24/7 without humans means you can't afford power issues. I'd oversize electrical capacity, backup systems, and network bandwidth massively. Future AI and robotics will only demand more power and data.

2. Sensor Networks & IoT Infrastructure - When you can't have humans watching, sensors become your eyes and ears. More sensors than you think you need, with redundancy, will save you from unexpected failures.

3. Manufacturing Execution System (MES) Software - This is the brain of your operation. Overspec the computing power, storage, and analytics capabilities. The software needs to handle way more data than you initially anticipate.

For underspec components, I'd intentionally go lighter on:

1. Fixed Robotics & Specialized Machinery - Technology evolves so fast. I'd use more modular, reconfigurable robots and avoid highly specialized equipment that locks me into specific processes.

2. Conveyor Systems & Fixed Material Handling - Instead of permanent conveyors everywhere, I'd design for flexible AGVs (Automated Guided Vehicles) and mobile robots that can be reprogrammed as needs change.

3. End-of-Arm Tooling - These are the easiest parts to swap out. I'd underspec here and plan for quick-change systems, knowing that product designs will evolve faster than my core infrastructure.

The philosophy here is simple: overspec the foundational elements that are hard to upgrade later (power, data, software), while underspecing the physical components that will inevitably need to change as technology and products evolve. This approach gives you the stability to run 24/7 while maintaining the agility to adapt to whatever comes next in manufacturing!