Most health apps fail the same way. You download them with good intentions, open them a few times, then quietly stop because logging anything requires navigating menus designed by someone who has never actually tried to remember what they ate for lunch two hours ago.
The app is always one more step than you want to take. So you don't take it.
I spent years cycling through the usual suspects. MyFitnessPal, Cronometer, and a brief and humiliating flirtation with a paper food journal. Every system worked fine for about ten days. Then life got busy, I skipped a day, and the streak broke. Once a streak breaks, the habit usually goes with it.
The problem was never motivation. It was friction.
The setup
About a week ago, I started using Poke, an AI assistant that lives in your phone's native Messages app. You text it the way you'd text anyone. No new app to open, no account to navigate, no loading screen. Just your regular messages thread.
I text Poke when I finish a meal. "Grilled salmon, brown rice, roasted broccoli." It parses the food, estimates calories and macros, and logs everything to Notion. I text it after a workout. "45-minute run, 4.2 miles, felt good." Same thing. Poke handles the interpretation, Notion gets the data.
That's the whole capture layer. A text message. It takes maybe eight seconds.
Poke connects natively to Notion, so I didn't configure anything beyond the initial setup. Every meal and every workout lands in its own database, timestamped and structured. From the outside, it looks like magic. Under the hood, it's just two tools that know how to talk to each other. I could probably replace Poke entirely at some point once I sit down and wire it up with a little Claude Code magic.
Making the data useful
Raw data in Notion is still just rows in a table. Useful to have, not useful to read. This is where Claude comes in.
I described what I wanted to Claude in plain language: a health dashboard that reads my Notion data and visualizes it. No code. Just a conversation about what I was trying to see. Within about 30 minutes, I had a working dashboard that pulls in everything Poke has logged and renders it in a way that actually tells me something.
Calorie totals by day, charted across the week. Macros broken down by fat, carbs, and protein, so I can see at a glance whether I'm hitting my targets or drifting. A scrollable meal log with timestamps. My exercise history with distance and duration. The dashboard refreshes automatically whenever new data is imported from Notion.
Looking at a chart of your week is a different experience from looking at a spreadsheet. The chart shows you the shape of your habits. You see that you eat well Monday through Wednesday, and then something happens Thursday. You see that your protein is consistently low. You see that you ran four times but skipped all strength work. Patterns you'd miss in a table become obvious in a chart.
The weekly coach
The feature I look forward to most runs on Sunday evening.
The dashboard reads seven days of data and generates a written summary of my week. Not a list of numbers. An actual narrative with a point of view. It tells me how my calories tracked against my goal, where my macros landed, how my exercise compared to the week before, and what it thinks I should adjust.
Some weeks, the summary is encouraging. Last week, it noted that I'd hit my protein target 5 out of 7 days, which is better than my average, and suggested I maintain that consistency heading into next week. The week before, it told me I was 400 calories over my daily average on days I skipped the gym, which is a pattern worth breaking, and recommended I plan a light workout for days when a full session isn't realistic.
It reads like a coach who has actually looked at your data rather than a generic app notification telling you to "keep it up." The difference matters. Specific feedback changes behavior. Cheerful noise doesn't.
When I replace Poke with my own custom-built text message logger, I might even add the personal coach there as well, so I can get real-time feedback as I log meals and workouts.
Meal planning and the grocery list
The part of the system that has done the most work is also the most mundane: the weekly menu.
Every Sunday, the dashboard builds a seven-day meal plan based on my calorie and macro targets. It accounts for what I tend to eat, balances the week so I'm not having the same thing three days in a row, and stays within the kind of effort level I've told it I'm willing to put in on a weeknight. Then it generates a grocery list organized by section, produce together, proteins together, and pantry items together.
I take that list to the store on Sunday afternoon. That's it. The week is planned. The food is in the house.
This sounds like a small thing. It isn't. The decision fatigue of figuring out what to eat every single day is a real drain, and it's where most diet attempts quietly collapse. At six o'clock, you're tired, nothing is planned, and you order takeout. Not because you're undisciplined, but because the path of least resistance won. Planning before shopping closes that gap. The easy choice becomes the right choice because you already made the hard choice at the grocery store.
What does this actually cost
Zero dollars beyond what I was already paying for Poke and Claude. No personal trainer, no meal delivery subscription, no expensive tracking hardware.
More to the point: I didn't hire a developer. I described what I wanted in plain language, and Claude built it. That's the part that still catches me off guard. The barrier between "I want a tool that does X" and "I have a tool that does X" was substantial. You needed to know how to code, or know someone who did, or pay someone to build it for you. Now you need to clearly describe what you want. That's a real skill, but it's a different kind, and most people already have it.
The friction is the feature
Every system I've tried before this one asked me to change my behavior in order to use it. Open this app. Navigate here. Log it this way. The implicit message was: the tool comes first, your life fits around it.
Poke inverts that. The tool lives in the place I already spend time. I don't open anything new. I don't build a new habit around an app. I already added one text to the conversation on my phone.
That's the whole trick. Reduce friction to near zero at the point of capture, and everything downstream, the dashboard, the coaching, the planning, gets to work with real data instead of the optimistic fiction of what you meant to log.