Hi HN — I’m a veteran founder working on something deeply personal:
helping people (including a lot of veterans and aging adults) stay on their feet longer.
We’re building PinkSteady — an app that uses the iPhone’s IMU data to estimate steadiness and deliver adaptive balance training in short daily sessions.
The problem:
Falls destroy independence, cost the healthcare system billions, and most “fall prevention” solutions exist only inside clinics. There are balance exercises online, but compliance is extremely low. We’re trying to make steadiness training simple + habitual + measurable.
Technical challenge:
Consumer-grade IMUs are noisy. Gait patterns vary wildly. Most people do not want to strap on sensors or buy devices, so we’re trying to extract useful low-frequency stability features from phones people already have — placed inconsistently in pockets, belts, hand, etc.
We’ve found:
– gravity vector drift matters
– window length selection changes user trust
– feedback timing affects behavior more than raw scores
– older adults ignore alerts but respond to gentle signals
We’re also experimenting with pink-noise audio and subtle cues (vibration or auditory scaffolding) to help people self-correct without scolding them.
What we’d love feedback on from HN:
– thoughts on fusing IMU signals in unconstrained device placement
– experiences with vestibular / gait signal modeling
– if anyone has insights on adherence mechanics in apps for older populations
– ethical considerations in giving people “scores” about their mobility
We just opened a small founding-member cohort to gather usability data and iterate.
Happy to share models, heuristics, failures, and learnings.
Jim Lucas, Co-founder of PinkSteady
We’re building PinkSteady — an app that uses the iPhone’s IMU data to estimate steadiness and deliver adaptive balance training in short daily sessions.
The problem: Falls destroy independence, cost the healthcare system billions, and most “fall prevention” solutions exist only inside clinics. There are balance exercises online, but compliance is extremely low. We’re trying to make steadiness training simple + habitual + measurable.
Technical challenge: Consumer-grade IMUs are noisy. Gait patterns vary wildly. Most people do not want to strap on sensors or buy devices, so we’re trying to extract useful low-frequency stability features from phones people already have — placed inconsistently in pockets, belts, hand, etc.
We’ve found:
– gravity vector drift matters – window length selection changes user trust – feedback timing affects behavior more than raw scores – older adults ignore alerts but respond to gentle signals
We’re also experimenting with pink-noise audio and subtle cues (vibration or auditory scaffolding) to help people self-correct without scolding them.
What we’d love feedback on from HN:
– thoughts on fusing IMU signals in unconstrained device placement – experiences with vestibular / gait signal modeling – if anyone has insights on adherence mechanics in apps for older populations – ethical considerations in giving people “scores” about their mobility
We just opened a small founding-member cohort to gather usability data and iterate.
Happy to share models, heuristics, failures, and learnings.
Jim Lucas, Co-founder of PinkSteady