Reliability

• Having experienced breaking changes with APIs, choosing to work with open and local systems instead (avoid_breaking_changes, 5)
• Avoid providers making changes to their online models behind the scenes that lobotomize the models (includes skeptics doubting this is actually happening) (avoid_inconsistency_model, 8)
• Having experienced bad service with APIs (avoid_sloppy_service, 5)
• Avoiding the liability of clouds failing due to natural disasters, war, regulation, etc. (avoid_catastrophic_failure, 2)
• Cloud compute is subject to surges, limits, and general unreliability compared to local compute (cloud_compute_unreliable, 2)
• The optimal configuration with which you use LLMs is dependent on your philosophy with the cloud (balanced take) (depends_on_philosophy, 1)
• If one has the skills, one can maintain a few self-contained frameworks forever (maintain_framework_forever, 1)