Home / Work / Motive
CASE STUDY · OUR OWN PRODUCT · PRIVACY ENGINEERING

One score hides
the contradiction.

Motive answers three questions about right now: what this moment is like, whether it is a good time for the thing you want to do, and if not, when. Most apps answer with a single score. Motive never averages: it keeps what supports and what opposes an activity as separate channels, and shows the contradiction as the finding. All of it is computed on the user’s own device.

Type
Our own product, live
Platforms
Web and Android
Core
Swiss Ephemeris in C, inference engine in WebAssembly
Privacy
Birth details never leave the device
Delivery
Every deploy verified byte for byte
Live at
motive.xors.workers.dev
A phone showing Motive's day view, held up against a starry desert night sky
0
Servers that see birth details
2,532
Live files checked on every deploy
2
Channels, never averaged
1,100+
Commits since August 2026
/ 01
The Problem

Apps that judge the quality of a moment almost always reduce it to one number. A single score averages away the thing that carries the meaning: when a moment is strongly favourable and strongly obstructive at once, a good time to analyse a deal and a bad time to sign it, the average describes no real state at all.

They also tend to want the most personal data a person has, their date, time and place of birth, sent to someone else’s server, and they speak in a jargon most people cannot check.

/ 02
The Approach

Support and opposition travel as separate channels from the astronomy all the way to the pixel. Ask about an activity and Motive says go, hold or wait, shows what pulls each way, explains why, and says when the window opens.

The computation runs where the data is: an astronomical core in C (Swiss Ephemeris) and an inference engine compiled to WebAssembly, in the browser or in the Android app. No server ever receives birth details.

The knowledge is authored from the classical texts, with each rule cited to its source and checked by a blind judge against those sources. The surface stays in plain language, with numbers the user can check and a “Why?” on every answer.

/ 03
The Hard Parts
  1. Never averaging, all the way through Keeping two channels apart is easy in one function and hard across a whole system. Every layer, from the planetary positions to the colours on the screen, had to carry support and opposition separately and report the tension rather than resolve it.
  2. Honest about what it claims Accuracy is defined as faithfulness to the sources, measured blind, and the product says so on its own face. It never claims proven prediction, and it asks users nothing about its own accuracy.
  3. Private by construction Moving the whole engine to the device meant shipping a scientific C library and the inference engine as WebAssembly, fast enough to recompute as the user moves through the day.
  4. Deploys you can prove Every release records the build’s fingerprint and then re-checks all 2,532 files of the live site against it, byte for byte. A mismatch is written down, investigated and closed, not ignored.
/ 04
The Outcome

Motive is live on the web and as an Android app, with a day view that shows the quality of each part of the day, activity cards that show the pull in each direction, and an Ask screen for any activity. The engine is in active development, with each change judged blind against its sources before it ships.

The architecture is the reusable part: compute on the device, verify every deploy, and never let a single number hide what the data actually says.

Motive's Now screen: the quality of the moment and a day timeline An activity card showing what supports and what opposes it The Ask screen listing activities and when they are favourable
Stack C → WebAssembly Swiss Ephemeris Next.js (static export) Flutter (Android) Cloudflare Workers Bun Blind source-fidelity judging Verified deploys

A single number reports the average
and describes no real state.

PRIVATE BY CONSTRUCTION

Need the computation to happen
where the data already is?

Contact Us →