90% retention + tiny wearables

Z / Runner 🏃🏻‍♂️

🏃🏻‍♂️ mapping out (n)stages to develop product(s) & services for the company.

• Dos: 1). Ai-native only 2). Useful to us at day ONE. 3). Distribution = MVP.

• Don'ts: 1). no hardware before PMF 2). Borrow heavy infra for as long as possible 3). No regulatory song & dance.

• Stages: 0). Own the problem: internally acknowledge that longevity sounds simple (live better → live longer), but walking down the path is hard: half-understood science, medical grade latin-english, hype, expensive biomarkers, yucky supplements, rainbow colored pills and protocols that feels like navy seal training camp. It is also an isolating experience, when family/friends think you're nut or weird. Huge cognitive load, reading all the books, watching/listening podcasts/videos, it is a full-time job. Realizing most of us don't have the right language to listen/communicate longevity & healthspan.

• (note: spotify just release 2025 top 50 podcasts, good to see Huberman lab is #9, rest are dominated by politics, sports, comedy, news).

• 1). AI-native app: "1 button, 60s, 7d/week" turn messy living into structured healthspan habits for n of n. UX: app alarm goes-off, 1 giant button, slide across 2x2 healthspan grid, press and hold to record/talk/prompt. (no typing, no journal, no modal hell, no graphs, no charts). Give instant micro-rewards (celebrate what you did, learned, and $won).

• App must down the unrealistic goal of 90% 30-day retention (explore basic utility like AI alarm clock?). App must have no empty state: first launch, app already knows your last 7 days from Apple/Google health, strava. App must not have cognitive load. There is never a question of "what do i write". It should be "press, talk, win". NOT notification, make it an alarm bell, a live ping. What NOT to build yet (no feed, social graph, comments, no long-form notes editor, chat agent. No onboarding essay or waxing poetry about what is longevity.

• Stage 1 is about shrinking longevity down to 1 button, 1 daily action in 1 app in 1 minute and let the AI do is job. Piggyback on the sensors and big-data on sleep, food, exercise. (Healthspan, dead simple). 2). Turn data to insights, actions → $rewards daily logs are noisy, AI compile/analyze "what changed" weekly scorecard and leaderboard for $reward, stack rank across cohort groups, so the w/w are always dynamic. Use tools like nano banana to generate "biographical memories", shareable on IG, YT, TK. We measure our success by % of users who share/post their "bios".

• Stage 2, AI "sherpa" if we have the right technical chops, can we fine-tune fast/cheap models into coach (i.e. LoRA). Simple step is to turn vetted advice, database, podcast shownotes, journals, books… as part of the db. Our key success metrics, will users PAY for the self-prompted expert coaching?

• Stage 3, Longevity Cafe, physical node… (see 12-4)

• 🏃🏻‍♂️1).does it make sense to have POV camera worn during a full-court full contact game? (i remember wearing even a lightweight, all plastic oakley frogskin sunglasses was hard, knocked off face, bouncing, sweats… and it weights only 18g). Point being, if the goal is to improve gameplay, then capturing external game-footage is more effective… bigger perspective on spacing, geometry, timing, schemes, defensive/offensive game habits. Not to mention, you want to have spatial context, off-ball movement that are outside of your POV (what did you NOT see during the live game).

• 🏃🏻‍♂️2). Case for a 2 camera side-line camera. If you were to have coaches w/ a clipboard by the sideline, which is currently "zero" for most of rec league games, that's a 10x jump and gives you an "unfair" advantage. Another sample, I have been following "shadowrun" (1 of the most popular long distance runner in Japan), his POV setup are ultra & go 3s. The value I get is his recording of other runners in his training camp, not his own running.

• 🏃🏻‍♂️ 3). Tl;dr 2 camera sideline setup (these are all magnetically mounted or clips), a). you will get better full-game data for analysis w/ your trainer and coach b). More meaningful off-ball, positioning, cut, shot selections, screening … c). You can film things you don't see

• 🏃🏻‍♂️ state-of-the-art = IBIS (in body image stabilization), software can only do much.

• 🏃🏻‍♂️ excluding all of the above, the field-of-view for human ~200 vs. the best gopro ultra camera is only 177. So you just won't have a good vantage view.

• 🏃🏻‍♂️ supporting argument for wearing a camera: if every player wore an ultra-tiny, or a chest bodycam (as shared above), then that's a game changer. imagine a 10 player feed, stitched & analyzed by AI, you can a multiperseptive 3D view. You can probably see what players like Magic Johnson "see & feel".

• 🏃🏻‍♂️ wow, you're definitely as bball sicko; my defaults: podcasts (no dunks on youtube), bill simmons, masterclass by curry.

• 🏃🏻‍♂️ my hot takes 1). NBA is liberal w/ remix b/c their partnership w/ Google (Pixel, YouTube, Curry…) 2). The scarcity w/ NBA is the star player, NFL is the team (shield)...

• 🏃🏻‍♂️ cost per episode is probably $10-$15k (assume B+ quality, (A grade = ESPN Sports Center)... profits for All the Smoke, The Pivot reportedly are making around $2-$5m per season (mostly sponsorship).

Stephen / Basketball 🏀

• 🏀 1. transformer-model native: language, speech, vision only. 2. useful to single-person or 1-peer play mode, even before 5 to 20 people social. 3. distribution via people who are on fire: lose 15 pounds, 1 red biomarker, upcoming boston qualifying / daily ymca games.
• 🏀 but the opposite is also true; see "simple rules" below.
• 🏀 what's your stage 0? Feeling the pain? Not having the right vocabulary?
• 🏀 why $ everywhere?
• 🏀 the 90% bar and the what-not filters are correct; the opposite thinking is: what are the values it provides every day – so much so that i will talk 60 seconds into and a few minutes reading / following it?
• 🏀 that's stage 1:5 (as a cohort of 5 people for social competitions) or 1:500 (as sharing to your 500 followers). what about the minimal of 1:1 – can our product be usable as in the solo mode? or, 1:2 – can a single peer be enough to boost the motivation, to clarify the confusing biomarkers, to spotcheck your whole-day health routine.. most of all: not to compete; see "divergence" below.
• 🏀 very possible and very scalable. let's start with a single thing you'd like to be coached on for coming days / advised on for more than a few hours. so we will try the 1-peer play mode "stage 1:2" above, but also maximize the existing ai tools / contexts (instead of waiting for a custom / fine-tuned model) to achieve that. for mine, what i need coaching now is…

🏀 #video-ai #tiny-wearable re: recording (basketball) film sessions – a headband with 6 cameras, a necklace with 10, or a collar clip / mindfly chest bodycam with 1 (for shooting drills)?

• z, which product on the market socially acceptable at the ymca for my games? insta360 go 3s ($429, 39 grams, 38 mins, 2.7k 30 fps) is 10x too big and too high res.
• d). Still pushing you to get a wearable, "Whoop"... you need to measure your recovery, kinetics… if it was good enough for lebron & kyrie,then why not give it a try?
• which off-the-shelf (ai) software that can stabilize despite jerky actions and can stitch together 8 views? adobe boris fx mocha pro planar tracker?
• sweat, slipping / dangling / choking, contact bruise.. make them unusable?

🏀 which is your favorite (nba) highlight channel? thinking basketball, jxmy highroller. (house of highlights) 🏀 loved hour-long podcasts last playoff, but wanted choices of condensed games to ease myself into the seasons. i found watching the netflix's "30 for 30" excellent, especially on time / reward.

• in a paragraph, describe how will you prefer your daily video cuts of nba games? 10 mins, top 5 plays (all camera angles: 15 for local, 20 for national, 55 for finals) or condensed close-score games, no spoilers in titles or thumbnails. [nba league pass offers condensed games, but..]
• nba is implicitly granting all creators to remix. thinking basketball and swish cultures are the official partners with 25k-hour archival access.
• what are the costs / profits of top creators? $10m/year for house of highlights (owned by warner bros), $500k for thinking basketball (70/30 split, via nba playmakers bbtv, by wsc sports ai).