Play It Forward — Life Horizon (the five-year question)

Status: plan / feature direction (2026-08-05). The long-horizon sibling of the shipped hours-scale Play It Forward forecast. Greenlit as a direction off the /play-forward typewriter hero (which already gestures at it: "…five years from now / when you look back on your life").

The one hard problem, stated first

The hours-scale forecast works because it predicts — caffeine's half-life, the sugar spike-then-crash — from a deterministic, source-backed knowledge base. Those are population averages; the tool can honestly say "this is what tends to happen."

Nobody can predict YOUR outcome — whether taking the job, moving cities, having the kid, or leaving the relationship makes you happier in five years. Any tool that claims to is lying, at the highest-stakes moments of a person's life. So the long-horizon version cannot be an individual-outcome predictor.

But there is a table — it just isn't a table of outcomes. It's a table of what people at the end of life consistently report regretting. That research is strikingly stable across sources, and it's the honest analog of the physiology KB: an aggregate of what humans reliably experience, used to ground the questions — never to predict the individual. See the next section.

The knowledge base that does exist: end-of-life regret research

The hours-version grounds in physiology averages. The life-version grounds in regret averages — decades of palliative-care and longitudinal well-being research that keeps landing on the same handful of things. Used the same honest way: source-tagged aggregate input to the reflection, never a claim about your specific future.

Bronnie Ware — The Top Five Regrets of the Dying (palliative nurse, thousands of deathbed conversations): (1) I wish I'd lived true to myself, not the life others expected; (2) I wish I hadn't worked so hard; (3) I wish I'd had the courage to express my feelings; (4) I wish I'd stayed in touch with my friends; (5) I wish I'd let myself be happier.
Harvard Study of Adult Development (the Grant Study, 85+ years, Waldinger/Vaillant): the strongest predictor of late-life health and happiness is the quality of close relationships — more than money, fame, or achievement.
Gilovich & Medvec — regrets of inaction dominate over time. Short-term we regret what we did; over a life, we overwhelmingly regret what we didn't do. Directly relevant to a "should I make the leap" decision.
Daniel Pink — The Power of Regret (19,000-person World Regret Survey): four core regrets — foundation (didn't do the steady work), boldness (didn't take the chance), moral (didn't do right), connection (let relationships drift).

And the positive mirror — not what people regret, but what people who lived long and well actually did:

Blue Zones (Dan Buettner / National Geographic — Okinawa, Sardinia, Ikaria, Nicoya, Loma Linda): the recurring "Power 9" — move naturally; sense of purpose (ikigai / plan de vida); downshift/stress relief; the 80% rule (hara hachi bu — stop at 80% full); a plant-slant diet; moderate, social wine; belong (community); loved ones first (close family); the right tribe (a social circle that makes the healthy thing the default). Notably: connection and purpose sit right alongside diet and movement — the same things the regret research surfaces.
Centenarian / octogenarian longevity studies (New England & Okinawa Centenarian Studies, Blue Zones interviews): across the people who make it far and stay well, the loudest common threads are strong social ties, a reason to get up in the morning, and moderate, consistent activity — not intensity, not wealth. Converges hard with the Harvard Study.

The through-line (why this is one instrument, not two): the hours-scale forecast already scores the daily inputs to exactly these outcomes — the soda, the drink, the skipped meal, the late night. Blue Zones + centenarian research is what those daily inputs sum to over decades. So Play It Forward is a single question — "what does this do to me, down the line?" — honestly grounded at each timescale: physiology for the next few hours, regret + longevity research for the next few decades. The near-term panel and the life-horizon mode are the same lens at two focal lengths.

How it's used (the honesty rule, identical to physiology): these ground the reflection questions and are shown with their source. Two flavors:

Regret frame — "People at the end of life most consistently regret not living true to themselves, and letting close relationships drift (Ware; Harvard Study). Of your paths, which best honors those — and which are you choosing to quiet someone else's expectations?"
Longevity frame — "The people who live longest and best share three things: close ties, a reason to get up, and moderate steady activity (Blue Zones; Centenarian Studies). Which path feeds those, and which quietly starves one?"

It cites, it asks, it does not predict that path A makes you happy or long-lived. The evidence sharpens the question; you still answer it.

What it is instead: a mirror, not an oracle

It never recommends. It does three honest things a person can't easily do alone in the moment:

Elicits your values — what actually matters to you (not a generic rubric): close relationships, health, money/security, meaning/craft, autonomy, growth, place, time. Drawn from what you've already told HiveJournal (journals, Downstream, stated priorities), confirmed and edited by you — never assumed.
Structures the tradeoff — lays each path you're weighing against those dimensions: what it feeds, what it costs, what it forecloses, what it keeps open. Not scored into a winner — surfaced so you can see the shape of the choice.
Asks the questions future-you would ask — in your own cloned voice. The regret-minimization frame the hero points at: not "will this make you happy" (unknowable) but "which of these does the you at the end of your life seem least likely to regret not choosing — given what you just told me matters?" You answer. It doesn't.

The output is not a verdict. It's a structured reflection + the hard questions + a saved decision record you can return to.

Why this is on-brand, not a pivot into life-coaching

Odessa north star ([../ai memory]): art and reflection that close the fiction→real-change loop without turning into coaching. This is a reflection instrument, not an advice engine — it hands the decision back to you, sharper.
Downstream's soul: the dread-vs-reality mirror. This extends it — you record what you predicted at the crossroads, and later the tool asks "you chose this; how does it feel now — closer to the version you feared or the one you hoped?" Over years that builds your own dataset of how your predictions track reality — the real skill isn't predicting a life, it's getting better at knowing yourself.
Your own voice: future-you asking, not an app advising. The person you're most likely to actually listen to is you.

The flow (v1 sketch)

Name the crossroads — two or more paths ("stay in this job" vs "take the offer" vs "go solo"). Free text.
Confirm what matters — the tool proposes your value dimensions (pulled from your journals/Downstream where available, generic starters otherwise); you weight/edit them. This step is the product — most bad decisions are values-blind, not information-poor.
Map the paths — for each path × each dimension: what it feeds / costs / forecloses / keeps open. You fill or accept prompts; nothing is fabricated about your life.
The far-future questions — 3–5 regret-minimization prompts in your own voice, grounded in the end-of-life regret research (each shows its source) and tuned to your top dimensions. You journal answers inline. The evidence sharpens the question; you answer it — the tool never picks.
Save the record — the crossroads, your values-at-the-time, the map, your answers, and (optionally) which way you leaned. Revisitable.
Later: the return — a scheduled check-back ("6 months on: how's the path you chose sitting?") that logs how it actually played out vs. what you predicted.

Reuse (mostly built)

A new life-wisdom knowledge base (life-horizon-core.ts, the honest analog of downstream-forward-core): a small, source-tagged table of the documented end-of-life regret patterns and the Blue Zones / centenarian longevity findings above, each mapped to the value dimensions it speaks to, so a prompt can cite the finding it draws on. Deterministic + golden-tested like its physiology sibling — the questions are grounded, the answer is always the user's.
The forecast UI shell + own-voice render (sayInUserVoice, jq_voice_clips).
Downstream value/priority signals + the dread-vs-reality calibration pattern (downstream-calibration-core) — the return-check is that loop at life-scale.
Journals as the values source; the query log pattern (migration 625) as the decision-record store.
Agreements/consent posture for anything weighty.

Guardrails (non-negotiable)

No prediction, ever. It structures and asks; it does not forecast life outcomes or say "you should." If it ever reads as advice, it's broken.
Counsel-gated seriousness. Major-life-decision territory borders mental health and crisis. Detect weight (leaving a marriage, self-harm-adjacent framing) → step out of the tool into real human/professional resources, never deeper into an AI reflection.
Values are the user's, confirmed. Never assume what matters to someone from their data without showing it and letting them correct it.
The record is private and portable. These are the most sensitive entries a person will ever write; honest-custody standard applies (user holds the key where encrypted).

Phasing

v1 — manual: name paths → confirm values → structured tradeoff map → own-voice regret-minimization prompts → saved record. No auto-inference; honest and shippable.
v2 — pull candidate values from journals/Downstream (confirmed, not assumed); the scheduled return-check + prediction-vs-reality calibration.
v3 — the meta-instrument: your personal track record of how your crossroads predictions land, so you learn your own biases (you over-fear change; you under-weight relationships; etc.). The point was never to predict a life — it's to know yourself.

Open questions

Naming: a distinct mode ("Crossroads") vs. a horizon toggle on Play It Forward.
How much to auto-draft the tradeoff map (LLM assist on prompts, never on your facts or a verdict) vs. keep it fully user-authored for honesty.
Where it lives: inside Downstream, or its own surface given the different weight.