[R] Which LLMs are actually best for bleeding-edge Linux/ML debugging workflows in 2026? [R] — PLINKFEED