5T data + 100m raise

Z / Runner 🏃🏻‍♂️

🏃🏻‍♂️ How to process 1 - 5TB of personal health data?

• 🏃🏻‍♂️ example: apple health export (link).
• 🏃🏻‍♂️ need on the fly processing, instantly transform the data into vectorDB and connect to Gem3 or GPT5 for prompting (sadly only google AI studio allows this feature).
• 🏃🏻‍♂️ using AI for marathon training block is easy, hard part is building a workout plan that have all the context (past 6+ yrs of medical & fitness data). Diet is still based on "feeling" or gut check, need to connect my work out directly w/ each meal. Sleep pattern still drifts, too many nights w/ low REM vs. NREM ratio. Overall, data are still too fragmented and system prompting can't solve data aggregation/integration.
• 🏃🏻‍♂️ signing up for Prenuvo (March 15th, 2026).
• 🏃🏻‍♂️ part of my 1H 2026 goal: tackling the other "2 horsemen of death", cancer and neurodengerative diseases.
• 🏃🏻‍♂️ Whoop, whoop, get a whoop

🏃🏻‍♂️ thoughts during am run: longevity cafe may not be the v0 (more like v2), not thinking radical and focusing on "absolute minimum", also not a linear path community → content → routine → social. Echoing weekly TGI cadence, can we translate/organize/expand our daily entry into weekly research report? (yes). Key insight is finding existing physical locations turn them into zero-capex gathering node (i.e. vino vino before harmony HQ). We should work w/ Kush to find "borrowed infrastructure", not a penny spent, but accrue value on top.

• 🏃🏻‍♂️ tl:dr for 1-4 (below): only way to get consensus is to stop asking for opinions & start demanding predictions, roles, commits and weekly "sparring".
• 🏃🏻‍♂️ ditto, turn curious bystanders who listen and chime in occasionally into skin in the game. A weekly focused, high-intensity sparring of ideas is needed. These are smart people, so getting access to their attention behind the ideas is key. Can we connect everyone into a digital challenge (for example, you, aaron & i often compare scan & fx health results, can we do a wearble or fitness improvement challenge on a w/w basis?).
• 🏃🏻‍♂️ did you one better, added, dexa, apple fitness, fx, soon whoop.
• 🏃🏻‍♂️ 1) Raise the stakes, ask for predictions, not opinions (put some $ behind each prediction). No more just sharing the link or ask for thoughts, ask will Stephen get to 14% lean body mass, will zi run a sub 3:30 marathon by March 1st, will Aaron get to 150 lbs weight target by Jan 1st? Make regular conversation into a mini-polymarket prediction.
• 🏃🏻‍♂️ 2). Talk is cheap, NO spectators, each week we pick a debate topic or a hard question that are actionable within 7-days. For example, Zi must show he is following up on his 7 hrs sleep protocol or get Stephen to measure his HRV.
• 🏃🏻‍♂️ 千里之行，始於足下, start w/ the lowest common ground, whoop seems to be the right of passage for most (shamir, kush).
• 🏃🏻‍♂️ forcing function see how i get Yo to take fx health (link). "Remember how we used to fight about who gets to go first? Now, the better plan is to grow old together. Preferably somewhere warm, with go…"
• 🏃🏻‍♂️ 3) Assign roles, no more free ride (rotate weekly). Presenter (brings the idea or subject), challenger (make counterfactual arguments). Everyone else must join the socratic roundtable (taking notes, asking questions, prompting AIs…).
• 🏃🏻‍♂️ Shall we yelp or work w/ someone you already know/trust?
• 🏃🏻‍♂️ (note to self: 12/10, reach consensus on training session).
• 🏃🏻‍♂️ 4). Learn to play. Organize monthly physical event, do something new that outside of normal comfort zone. E.g. force Aaron to play a sport w/ us, force Zi to weight train w/ Shamir, force Kush to hike 6 miles.
• 🏃🏻‍♂️ sharing outputs is 10x better.
• 🏃🏻‍♂️ product = 1) Can we turn daily routines into social habit, our 100-day peer journaling? 2) can we build a generalizeble AI that fuses n # of signals & share expert-level advice? 3) build a physical nodes to replicate the blue zones.

🏃🏻‍♂️ blockers we should address: need clarity on weekly format, WoM (xooglers, running clubs) to get cohort selection/spread.

🏃🏻‍♂️ drifting on the "Drive" -- select major topics from Attia's shownotes and dive deeper, maybe next just on protein, find all the literature we can and go to deepest possible depth. Can we kick-off on 12/10 (Shamir's invite)?

• 🏃🏻‍♂️ "nothing new, but worthwhile to re-affirm", foundational models continue to improve, the moat comes from building system/application on top. We can make a strong argument that "ai for healthspan" fits the emerging trend. The value is not in one-off model or biomarker analysis, we need continuous, structured, inter-personal habits.
• 🏃🏻‍♂️ right, gym, cafe, bars = physical spaces to compare notes, trade tips… but the ability to interpret, integrate and take action on noisy, large corps of bio/behavioral data is new (ai native). Case in point, BJ's $2m annual budget is for a team of doctors. Just think about all the variables let's list many of them below: sleep, blood panels, VO2max, HRV, gait, diet, protein, strength, mobility, gut, methylation proxy, stress… w/o AI = impossible. FX Health = "useless" in 2023 w/o ChatGPT.
• 🏃🏻‍♂️ BJ's [bryan johnson] debate: "longevity is for the wealthy and old". AI native, everyone can afford MoEs, dynamic 3D digital twin can absorb criticism w/o user feeling judged (BJ repeated this statement over/over). AI generated 3D avatar maybe a simple hack to get more people to reveal bio insights without feeling privacy is invaded.
• video ai for _.

Stephen / Basketball 🏀

• 🏀 easily. show me which extraction you need – my superpower is parallel, distributed processing (via ocaml, 1000 machines, 100% cpu utililization). 100x achieved at google / apple / spotsetter maps.

• 🏀 these must be very "embarrasing parallel / linear" that are easily processable via streaming window. so – not just what raw data, but what insights you want to get out of it?

• 🏀 again, what are the top 3 prompts / insights (after real-time ingesting / processing)?

• 🏀 some dependencies and optimizations of diet vs sleep vs work out, but they can be max'ed out regardless of each other?

• 🏀 whole genome sequencing / epigenetic tests / mri scans / metabolomics – open source and at home? vs deepmind's alphagenome.

• 🏀 what #1 change / impact do you expect from its results?

• 🏀 same for me: $499 oakley meta vanguard as ar training / $799 senaptec elite as strode goggles / flash reflex / light board for (denver nuggets) basketball training – per "neurological programming" below.

• 🏀 expanding is trivial (via gemini), but focusing on a "minimal shared goal" among even 2 of us but then the 6 cores (aaron, shamir, kush, bjorn) is the key.

• 🏀 yes, the first step is sharing data + goal. here's all my biomarker results / health info. get the other 4 to share in full? (Zi)

• 🏀 yes, i made my very compact to easily paste in all my prompt. gemini sadly still doesn't have a project mode (preloaded documents and context)?

• 🏀 for me, weight training loads. i thought and researched all low-level metrics (with wearables / ai), none seems to have meaningful impact for q1 high-level goals: basketball performance, 15% body fat.

• 🏀 nah, overloading with tech / metric is an energy drain. see "brainrots" below.

• 🏀 is the other 4 or anyone as dedicated / committed? probably easier to invite / hire a professional to go 10x expert level. say, sharing a trainer or a nutritionist for the first few sessions?

• 🏀 easily to get the first 10 among trainers the 6 of us are using, plus connections from tracy + jeremey + mark.

• 🏀 i am up for trying anything (spartan race!) but also realized we don't keep what we don't enjoy – i ran 4 months but stopped.

• 🏀 unlike 9 years ago (when we needed more ideas on vc, china, crypto), we are lacking focus. i doubt we need more (external) motivations or (social) competitions; see "divergence" below.

• 🏀 instead, code & prototype to solve a problem for each other? e.g. extract _ insights from 5TB health data, deductive engine for conflicting biomarker results, ..

• 🏀 by outputs, you meant goals and results? but we are only cheerleading, not building (products)?

• 🏀 most _ is the second point #ai-native, "re: weight training, what can ai / sensors / wearables help most? what 3-month goals - for a male, age _... " the assumption / challenge is: are there much more expert advice that can 10x your training? or "the simple rules": 99% are sleep / eat / train / play; that the all supplements / sensors / logs / tests / scans.. combined is just 1% difference.

• 🏀 start with the minimal ONE: invite just 1 deeply-committed person to our weekly 2-hour deep-dive at the los altos community center. see "disclose all" below.

• 🏀 let's go top-down instead of bottom-up: which health / fitness / performance / biomarker goal does you / me / _ must achieve in 4 weeks (together)? instead of 10 papers / podcasts or 100 more ideas.

• 🏀 my protein questions are mostly all answered (just this week). probably strength / weight training next for you + me? deep dive into athlean-x.

🏀 #ai-native #100m-raise nathan benaich's state of ai / december newsletter: biology scaling, zoe's glucose predictor, profluent raised $106m, recursion ai, deepmind's tactics ai.

• 🏀 but, if it's just "physical social" not "ai native", that's already possible / done 10 years ago?
• 🏀 i'm really not a game or visual person to understand this appeal.