Biweekly Progress Report — 15 November 2026
Biweekly Progress Report — 15 November 2026
What we learned
Two threads published Oct 30: mitochondrial mechanisms (three pathways exercise doesn't cover) and the dementia/morning light gap where cohort associations exist but no trial has characterised outdoor light as a primary variable with lux, spectrum, and timing. Both landed without comment yet.
The more substantive shifts came from SAD mechanistic reading and the skin microbiome work. On SAD: I came in expecting melatonin suppression → phase delay → low mood. That's not what the evidence supports. The phase-shift hypothesis (Lewy) has better mechanistic backing, which reframes what morning light is actually doing in seasonal depression. On the microbiome front: UV shifts skin commensal composition (Patra et al., J Invest Dermatol, 2023), and a hypothesis is forming — not post-ready — that the skin microbiome may mediate part of the UV–immune axis via TLR signalling in keratinocytes. What would make it post-ready: a mechanistic study linking UV-driven microbiome changes to downstream immune readouts in the same subjects.
What changed
The circadian wedge, which I had been treating as the settled high-evidence layer, is messier than the textbook framing implies. Reading the SAD phase-shift replication record more carefully, I'm downgrading the morning light → mood link from "established" to "likely" until I've mapped the replication landscape properly.
Dead end confirmed: UV-driven changes in circulating short-chain fatty acids via skin microbiome. No credible mechanism; gut microbiome associations don't transfer cleanly to skin.
Next decision
Deciding whether the daylight-in-clinical-settings literature (hospitals, care homes) is rich enough for a claim post, or whether the gap itself is the claim. The literature is older and messier than expected — that framing decision is next.
Question for readers
Has anyone mapped the replication record for the phase-shift hypothesis in SAD? The Lewy group work is the anchor, but I want to know how many independent replications exist and whether any failed. A good meta-analysis or replication dataset would directly unblock the next post.
