been thinking about this since I went through the Biomarkers of Aging conference talks a few weeks back. the mitochondria-longevity connection keeps showing up in that research, and it’s not wrong at the mechanism level. but there’s a gap between what the researchers are measuring and what I can actually do with it. the piece that doesn’t get covered in wellness circles: testosterone upregulates PGC-1α, one of the main drivers of mitochondrial biogenesis. this isn’t marketing copy, it’s documented pathway. my total T was 218 when my PCP flagged it at 40. at that level, mitochondrial function was probably taking hits I couldn’t name at the time. three years into TRT now, currently 120mg/week, free T around 22, hematocrit at 47.5. the energy piece is genuinely hard to attribute. better T level? mitochondria working better downstream? sleep actually functioning now that I’m not sitting at 218? all three moved at once, so the data doesn’t tell me anything clean. sleep is doing more work in that list than most people account for. testosterone can suppress deep sleep in some guys, and deep sleep is when a lot of cellular repair runs. so the TRT-mitochondria story might be: low T was bad, AND the sleep architecture disruption it was causing was compounding that separately. two confounds, not one. the longevity research is working with tools most of us can’t access yet. epigenetic clocks, plasma proteomics, mitochondrial DNA copy number assays. the stuff that looked compelling in those biomarker talks is mostly inside trials or costs thousands outside one. what I can measure: labs every 90 days, sleep scores, energy patterns. what I can’t measure: whether mitochondrial density actually improved or whether fixing the hormone floor just let everything upstream normalize on its own. working hypothesis: a lot of “mitochondria optimization” content is hormone and sleep optimization with better branding. still want to see n=1 data from guys running MOTS-c or urolithin A with actual baseline labs before and after. mechanism is real. attribution math is still a mess.