Sentient OS – On-device intelligence layer for your entire digital life
Custom vision LLM runs overnight on phones when cloud AI violates privacy.
Offline ski analysis skips cloud and subscriptions—sensor fusion without the yearly fees.
Skiers and snowboarders wanting performance analytics without subscriptions or data privacy concerns.
Strava (sports tracking) · Slopes (ski app) · Garmin ski/snowboard metrics
No cloud processing. No account system. No subscription model.
I originally built it because I was frustrated with two things: 1. seasonal sports apps charging year-round subscriptions 2. performance data being tied to cloud accounts for no strong technical reason
Instead of debating the model, I decided to try building something different.
⸻
The technical approach
The app processes: • GPS • motion sensors • acceleration data
All analysis happens on-device.
The biggest challenges weren’t UI. They were: • sensor noise during aggressive turns • battery drain during long ski days • balancing sampling rate vs. usable signal • designing a scoring model that feels intuitive but is still technically grounded
The offline requirement made architecture simpler in some ways and harder in others.
No backend meant: • no server costs • no sync logic • no auth system • no cloud ML pipeline
But it also meant: • everything must run within phone constraints • optimization matters • no ability to “fix it in the cloud”
⸻
What didn’t work • I overestimated how much people care about privacy as a primary selling point. • Distribution is significantly harder than engineering. • Launch posts don’t move the needle nearly as much as you think.
The app currently has 600+ downloads across iOS and Android. No investors. No paid marketing.
Getting the first 100 users was easier than getting from 300 to 600.
⸻
The business decision
I chose one-time pricing instead of subscription.
From a pure revenue perspective, subscription would likely generate more predictable income.
But I wanted to see whether a niche sports product could survive without recurring revenue and without cloud dependency.
It may turn out that this was financially naive. I genuinely don’t know yet.
⸻
What I’m struggling with now • distribution without hype cycles • explaining “offline” in a way that isn’t marketing-sounding • deciding whether staying small and sustainable is enough
I’m curious how other builders think about: • offline-first architectures in 2026 • one-time pricing viability • building for niche seasonal markets
Happy to answer technical questions.
Custom vision LLM runs overnight on phones when cloud AI violates privacy.
Local-first running tracker that ditches subscriptions and cloud sync for pure on-device privacy.
Free local dictation that undercuts Wispr Flow and Superwhisper on price.
Hardware + local inference + P2P, but ships March 2026 with zero proof of technology working.
Beats robot_localization on 5 of 6 benchmarks with zero manual tuning.
Free local dictation when Superwhisper and Wispr Flow already charge monthly.