Launching today
Illume Labs
24/7 personalized health companion you can text
34 followers
24/7 personalized health companion you can text
34 followers
Illume is a 24/7 personal health companion that connects your wearables, nutrition, workouts, & lab results to give you personalized, actionable insights over text. Sign up: www.illumelabs.ai Illume connects patterns across your data that individual platforms can’t see, giving you a more complete picture of your health instead of another isolated score or dashboard. You can text your Illume agent for quick answers or use the platform to explore your data, trends, and history in depth.
Illume Labs
@pari_latawa The proactive insights sound really interesting. How do you make sure the recommendations stay personalized and aren't just generic health advice as more user data comes in?
Illume Labs
@vishnu_kant_07 Each recommendation is generated from each user's own longitudinal data (connected wearables, blood test results, & their food and fitness logs). As more data comes in, we can build a better picture of their baseline, ID changes or patterns, & suggest actions based on specific goals and health trends!
Pari, the lab upload is the part I keep coming back to. Sleep and recovery questions have no wrong answer that hurts anyone. Bloodwork does, and people usually see a result before a doctor has explained it, so they arrive at your text thread already worried.
The design problem there is not the insight, it is the refusal. At 2am, with a value genuinely out of range, saying nothing is an answer and sounding calm is also an answer. Building clinical AI, what earned trust was the system being explicit about what it would not do, not how good the output was.
Is there a defined escalation path for out of range results, or does the model decide case by case?
Refocus
The cross-source pattern detection is what stands out to me. Most tools silo wearable data and lab results separately, but the "ApoB dropped, hs-CRP jumped, worth a look" example in your screenshots is exactly the connective insight people miss on their own. I build voice AI that checks in on aging parents daily, so I think a lot about turning passive health signals into something a person will actually act on. How do you decide when to proactively surface an insight vs. waiting to be asked? Getting that cadence right without nagging feels like the hard part.
Illume Labs
@igorgurovich We're doing morning/nightly briefs on health activity as a baseline, but we can answer any questions a user has at any time. Curious what you've found to work well?