i think it's safe to say G2/G3/G4 are out. not sure if i said that before. this is no longer a "scenario." they are out. we half our total config possibilities.
covariates are not out of the picture. but they should be contingencies. as for how many covariates, they are prob all going to be manually assigned if necessary. starting with col 11 and would possibly extend as far as col 9 (in the direction of 11 to 1). not sure if necessary. but a manual contingency based on area average across columns on day 47 (final morphology before freezing).
and since MAP2 is out of the picture per beatriz's staining on block 21 (all negative), we will be left with 5 biomarkers: CTIP2, TBR2, TBR1, HOPX, SOX2.
per chatgpt 5.5 pro deep research:
- CTIP2 would be the most promising.
- TBR1 second
- TBR2/SOX2 third
- HOPX kinda weakass, but keep
also per gpt 5.5 pro on absolute vs. normalized quantifications, which involves DAPI:
for the primary run, likely finalized, as of May 4, 2026 at 12:36 AM:
total configs: 5 × 2 × 7 × 47 = 3,290
fits/config: 25 outer fits × 5 inner folds × 600 hyperparameter combos = 75,000
total fits: 3,290 × 75,000 = 246,750,000
- likely fucked up the sosv sf meeting w/ mohan (gp) and phil (mohan's guy) after referral from ned maybe due to my brain being fried. they told me to introduce myself and i fucking went off the rails talking about self actualization and aging philosophy, and gotten same response as from sam strickberger. idk if this is a good sign or not. they left me homeworks and so i spent 8 hours on friday (from 12 to 8 pm that day) to write a follow up to them answering customer conversion on the basis that they don't trust us even if the reproducibility is not a problem but organoids in general, and a assumptive projection of venture revenue scaling to 100M. first demand of answer is from phil, second from mohan. no reply from them yet but that's what they said i need to do as homework. so ig i still have a chance to salvage. i learned they don't care about the technical themselves. i emailed ned earlier today also after realizing i wanted to figure out their fit for telora, not the other way around as the primary purpose of my reach out, but not accusatory.
- as funding pathway contingency i tried reaching out to jack (amherst alum, semiconductor startup CEO backed by pillar/petri), no response yet. emailed yesterday.
- beatriz is not sharing iCorps meeting notes/info with me. "i don't feel super comfortable sharing with me" oh well. so much about working together.
- meeting tmr at 1pm with kelly (and possibly more ppl at histospring) to figure out how to do the 3d proxy aka 5-panel multiplex staining for the target biomarkers for all organoids therefore blocks, downstream of sectioning, upstream of imaging + quantification. more notes about the ground truth collection as written:
idk how things will go from here. one thing at a time. i need to sleep, wake up normally starting tomorrow, eat well, go to the meeting with a clear mind, not go off track, and solve one blocker at a time.
planning with kelly and histospring.
involve jim chambers.
ship the actual blocks, and an additional test block for jen.
jen does sectioning on 24 blocks. kelly (or someone else at histospring) covers the staining properly.
imaging. quantification. modeling.
if good, call suzanne.
sosv sf/ny would be in parallel. idk how dead im to them, but i will sure be more alive if my PoC works at all.
rest. nothing is more important than waking up with a clear mind. go outside. maintain high recovery score. eat well. don't ruminate. just go outside for a walk if i have to. remember what nieztsche said: "Never trust a thought that occurs to you indoors."
for some motivation booster, per dan said directly to me: "you are both formidable and earnest."
quoted context from yc: "when founders are both formidable and earnest, they're as close to unstoppable as you get."