THE LONG
GAME.
This is the same bar I bring to product work, applied to my own life and my own data.
[EIGHT YEARS OF DATA]
Eight-plus years on the Apple Watch, through several upgrades, with one on most members of my family. The three rings close every single day. Sickness, travel, time-zone shifts, none of it has broken the streak.
The data piled up faster than I had stories to put around it. Resting heart rate, HRV, sleep stages, recovery, training load. The Activity app could read its own metrics back to me and that was about it. I just wanted to understand where I am, inside and out.
Cycling runs across gravel and road; I keep equal time on both. I lift heavy on a fixed schedule and run zone-two cardio for the conversational-pace work. Meditation and yoga go back four or five years. The training spreadsheet rides with me to the gym every month, repetition by repetition. Gratitude journaling started about a year ago.
Longevity, for me, is not adding years for the sake of adding years. The practices above, with eating and sleeping underneath them, are aimed at being stronger inside the life I already have, and at carrying the things I love into the decades ahead with more presence than my body would manage without them.
A heavy lift, a long ride through dry hills, a quiet hour in the woods each return something nothing else does. Walking in nature, a real meditation session, gratitude journaled out loud all belong in the same tier. Mindful practice ranks alongside lifting and cycling for that reason.
[THE FIRST STACK]
The first version of my longevity stack was an app I built on top of Apple Shortcuts and posted on LinkedIn in December 2025. Shortcuts pulled what they could from HealthKit, posted to a Vercel function, stored the result in Redis, and Claude reasoned through the data via MCP. I would say "GM" and Claude would tell me how my body actually felt and what to watch. The deploy took seconds, which reminded me why developer experience matters. The open-source half of that work is on GitHub.
About a year into gratitude journaling I switched from writing to voice. The richer signal came through immediately. Voice carries cadence, intonation, and the kind of detail my fingers could not match at the speed I think, and the practice survived because of it. Multimodality is the bet I am making for the next decade. Voice, touch, feel, hearing, every input one level closer to how I actually live and work is the right input to the agent layer that runs on top of it.
Apple Shortcuts could not reach the bulk of HealthKit. The richer signal sat behind the SDK boundary, gated by a developer account. I built Prana to get past that gate.
[PRANA · BODHI]
Prana is the native rebuild. Swift 6, SwiftUI, iOS 26, with HealthKit background delivery permissions across the breadth of Apple Watch metrics. It runs on my own developer account, on my phone, encrypted, my data alone. It does not ship to the App Store. Apple's on-device model handles voice journal transcription locally. The data flows through a Vercel serverless backend (Python, Fluid Compute) into Upstash Redis and through MCP to Hermes, the Mac Mini I bought specifically to host this work.
Bodhi runs on top of Hermes through SOUL.md, currently powered by Claude Sonnet 4.6 for morning-brief reasoning. The architectural pattern is a council of LLMs. Local models like Mistral, Kimi, and GLM handle routine routing and orchestration, with Claude Opus over the API for the heavy reasoning passes. Bodhi decides which model fits the work in front of it.
A cron at 6:15 in the morning fires a Bodhi skill that pulls the latest week of HealthKit data, the gratitude entry from the night or morning before, and historical briefing context. The brief composes and lands on a Telegram bot. I read it before I open anything else.
The brief illuminates patterns I would not have caught on my own. It tells me what HRV is doing and what is likely driving it, whether mouth taping is working or needs a different brand or bedtime, whether the dinner-to-bed window is too tight for the sleep score I want, and whether the last seven days of HRV, resting heart rate, recovery, and zone-two minutes are pointing at over-training or under-training. Zone-two trends are tracked week over week, with a weighted-vest prescription when the curve goes flat, a modality swap when one is called for, and a rest-day call when the gym is the wrong move. Most mornings the brief surfaces the drift before I notice it myself.
[FORWARD]
OpenClaw, Hermes, the new Prana, and Bodhi as the persona all came together in February 2026. Mac Minis have been short on stock for months. Everyone serious about personal-agent infrastructure is buying one. I wanted a front-seat experience of that, which is why I took this path.
The longer arc is Bodhi as autonomous life-operator. The same orchestrator that composes a morning brief grows into a never-ending while-loop handling scheduling, planning, renewals, emails, memberships, dues, and any other proactive task currently consuming my attention. The named-agent fleet underneath Bodhi is the same one I describe in the Paxora portfolio: each function I ran by hand during Paxora's first year is becoming a sub-agent in this layer, and Bodhi orchestrates the lot across the personal surface and the business surface together. That layer has not shipped yet. The work between now and the end of 2026 is to bring it up.
By the end of 2026 I expect personal AI assistance to commoditize. Watches, sunglasses with cameras, AirPods with cameras, every wearable streaming richer signal closer to the heartbeat. What stays scarce is human presence, and the time in nature without a screen mediating any of it. I built Bodhi to protect that human time while the rest of the loop runs on its own.
The morning brief assembles itself. I keep building the system underneath it. Each cycle, I engineer myself out a little more, so the human time the loop is meant to protect actually gets used.
