Tether

Inspiration

Habitual overspending is often treated like a willpower problem, but the research says it follows predictable patterns. Clinical evidence around compulsive buying shows a repeatable cycle (anticipation → preparation → shopping → spending), with urges leading to purchases most of the time. We built Tether around that insight: if behavior is predictable, financial protection can be proactive instead of post-mortem.

What it does

Tether is a mobile-first, sensor-aware financial agent that sits between your environment and your money decisions. It combines location dwell signals (like spending too long in a high-risk retail zone) and physical impulse signals (like stress-scrolling patterns) to estimate risk in real time. In Advisory Mode, it gives contextual prompts; in opt-in Execution Mode, it can run predefined actions like liquidity balancing, temporary wallet/card lock, and intervention prompts with clear user override.

How we built it

We designed Tether as an on-device pipeline: sensor fusion (location + motion) feeds a decision engine that moves from advisor to operator based on user settings. The product logic is defined by autonomous boundaries (risk tolerance, transfer limits, cooldown/override rules), and every action writes a local Decision Memo for auditability. The demo direction includes geofence-triggered nudges, price-match interventions, planning recommendations, and opt-out blocking controls.

Challenges we ran into

The hardest part is balancing automation with trust. Sensor signals are noisy, so reducing false positives while still acting in time is a core challenge. We also had to design autonomy boundaries carefully so the app can act decisively without removing user control.

Accomplishments that we're proud of

We turned a reactive budgeting concept into a proactive, behavior-aware agent architecture. We built around data sovereignty (on-device processing, no raw sensor export), and added explainability through Decision Memos so users can always see trigger, action, and why. We also established a clear product wedge: sensor + transaction correlation that traditional bank apps do not capture.

What we learned

Behavioral finance products live or die on trust, not just model quality. “Quiet most of the time, decisive when it matters” is a better UX principle than constant alerts. We also learned that autonomy needs explicit boundaries and easy override to feel protective instead of intrusive.

What's next for Tether

Next we’re focused on shipping the demo pillars as production features:

submission to another program After that, we’ll deepen controls (limits, safe mode), improve action reliability, and measure against our core outcome: reducing unplanned retail spend in the first month.