Loga: Find Your Purpose
The Problem
Most goal and habit apps ask “what do you want to do?” and stop there. Check a box, keep a streak, feel briefly good. But that framing skips the harder, more important question: why does this goal matter in the first place?
Without that layer, habit trackers tend to produce the same failure pattern — a burst of motivation, a few weeks of checked boxes, and then a quiet abandonment once the novelty wears off. The habit was never anchored to anything deeper than the habit itself.
Loga starts from a different premise: sustainable behavior change requires meaning underneath it. The app is built on frameworks from thinkers like Viktor Frankl, Jordan Peterson, BJ Fogg, and Aristotle — translating ideas about purpose and identity into a product people actually use day to day.
Who It’s For
Loga is built for people like Jack, 29 — not in crisis, not chasing a dramatic transformation, but stuck at a quiet plateau. Stable job, stable life, and a nagging sense underneath it of “I’m fine, but what is any of this actually for?”
Jack tried conventional habit trackers before Loga and bounced off them. Checking a box for “drink water” felt disconnected from anything he actually cared about. He needed a tool that started with why before asking what — something that could hold both the abstract (what do I value, who do I want to be) and the concrete (what do I actually do on a Tuesday).
Key Design Decisions
1. A three-layer architecture: meaning → behavior → feedback
Rather than one flat list of goals, Loga is structured in three layers that mirror how real change actually happens:
- Meaning discovery — helping someone articulate what they actually value, not just what they think they should be doing
- Behavior design — translating that meaning into specific, achievable actions (goals, milestones)
- Feedback — journaling and reflection that closes the loop back to meaning
This structure shows up directly in the app’s navigation: Dashboard, AI Coach, and Journal are all user-facing expressions of this loop.
2. An invisible context layer
Early versions of Loga included a “Context” page — a place where users could see and edit what the AI understood about them. In testing, this created the wrong dynamic: it turned self-understanding into a data-entry chore.
The fix was to remove Context from user-facing navigation entirely and let it run invisibly in the background, built automatically by the AI as it observes goals, journal entries, and coaching conversations — functioning more like a therapist’s private notes than a profile the user manages. This was one of the most significant pivots in the whole build: a feature engineered to feel absent, because its presence undermined the experience it was meant to support.
3. An AI Coach with real memory and real actions
The AI Coach isn’t a generic chatbot bolted onto the app — it’s built on a two-layer prompt system (a state layer that tracks where the user is, and a routing layer that decides how to respond) with 36 distinct prompts tuned to different coaching moments. It can take real actions inside a conversation — creating a goal, updating a milestone — surfaced to the user as confirmation cards they approve before anything changes.
Model selection is deliberately split: a fast, cheap model classifies intent and routes the conversation, while a stronger model handles the actual coaching exchange — balancing response quality against real API costs at scale.
4. State caching over real-time computation
Every AI interaction could theoretically recompute a user’s full context from scratch — but that’s slow and expensive at scale. Instead, Loga caches computed state and only recalculates it when something meaningful changes (a new goal, a completed milestone, a significant journal entry), rather than on every session open. A deliberate tradeoff of a small amount of staleness for a large reduction in cost and latency.
Outcome
Loga shipped as a complete MVP: four core screens, a working AI coaching system with tool-use and confirmation flows, ten Supabase tables with row-level security, subscription infrastructure via Superwall, and Sign in with Apple — built from a standing start with no prior programming experience.
The bigger outcome is the product thesis itself: that meaning and behavior aren’t separate problems requiring separate apps, and that an AI coach can hold both without becoming either a therapist replacement or a glorified to-do list.
Reflections
Building Loga solo meant every design decision had to be argued for and defended without a team to pressure-test it in the room — which made the invisible-context-layer pivot, in particular, a genuinely hard call to make alone. It’s also the project that taught me the most about the gap between a screen that looks finished and a screen that’s actually considered — spacing, hierarchy, and restraint in service of a genuinely different idea about what a “productivity” app can be.