Arnav Chakravarthy

20 · math + bioe @stanford · ai @hundredhealth · prev. ceo @sciLynkhq 🧑‍🔬

[Arnav Chakravarthy @arnavncx]

I’m beyond excited to announce @SciLynkHQ has raised $100,000 from @1517fund! Huge thanks to @DStrachman and the 1517 team for believing in us. On behalf of the SciLynk team, I would like to give immense appreciation to the following individuals and groups:

[Danielle Strachman 💗 🐈 💃 🪴 🎸 🎨 🐕 @DStrachman]

Michael and I were on the founding team of the Thiel Fellowship and left to scale our work backing young ambitious minds and builders before anyone else.

1517 turns 12 next year. And though we fly humbly under the radar, you know some of our greats: Lambda, Loom, Mach, Positron, ntop, Deepgram, Rainmaker.

As of last week, Fund I is at a 5X DPI. We were the first check before anyone else would have a serious conversation with these founders and by doing that, we (a teacher and philosopher) have turned $20M into over $100M which doesn't even include our SPVs to continue driving our wins home.

There is nothing like building a small community that you believe in before anyone else, you see something special and cultivate the hell out of it, and then grow that into a force of nature with multiplying effects.

Best is yet to come!

[Arnav Chakravarthy @arnavncx]

your CGM catches a glucose spike at 2am. your doctor's dashboard never updates. three days later they're staring at last week's averages when you walk in. append-only breaks the second patients expect real-time care instead of retrospective guesswork

[Arnav Chakravarthy @arnavncx]

@agingdoc1 circadian alignment or just correlation with early risers having better sleep hygiene overall? would love to see the confounders they controlled for

[Arnav Chakravarthy @arnavncx]

@wallstengine readiness score is sick but the killer feature is actually explaining what to DO with it. most wearables dump HRV numbers with zero context. if apple ships clinical reasoning like 'your score's 65, skip heavy lifting today', thats actually cool

[Arnav Chakravarthy @arnavncx]

@SuccinctLabs @OpenAI does verifying the compute trace actually tell you anything about alignment or safety though? like proving 'we ran this training run' doesn't prove 'the model won't do X', right?

[Arnav Chakravarthy @arnavncx]

@stanleyzhaoo like we never left

[Arnav Chakravarthy @arnavncx]

A few months ago I opened scilynk, the startup @stanleyzhaoo and i worked on in high school and immediately felt the itch to refactor the whole (1/7)

[Arnav Chakravarthy @arnavncx]

@stanleyzhaoo Honestly just felt good to ship something real again to a project that meant a lot to me... even if it's a product i don't actively work on anymore, seeing researchers still get value from it made the rebuild worth it (7/7)

[Arnav Chakravarthy @arnavncx]

@stanleyzhaoo Whole thing took less than a week since cursor let me move like 1000x faster than our pace for the original build. Same app, same users, but it finally feels like it was built in 2026. (6/7)

[Arnav Chakravarthy @arnavncx]

@willdepue the behind-the-scenes frontier lab jousting is probably wilder than whatever we'll see publicly. these orgs are literally racing for AGI, every capability leak or talent move gets weaponized. we'll get OpenAI's version but it'll be sanitized

[Arnav Chakravarthy @arnavncx]

@Parsifaler curious what biomarkers you're tracking to quantify the aging acceleration. is it methylation clocks, telomere attrition, inflammatory panels, or something else?

[Arnav Chakravarthy @arnavncx]

@Apriyadi_sardi nvidia literally prints the chips that train the models. of course Jensen's calling it first, he sees the compute scaling before OpenAI writes the blog post. same reason TSMC knew iPhone demand 6 months before launch

[Arnav Chakravarthy @arnavncx]

@robertchiima warriors haram basketball is genuinely the most accurate description of our offense i've heard all year 😭😭 we run the same steph gravity play 40 times and pray someone hits a contested three

[Arnav Chakravarthy @arnavncx]

@SmartTechTrendz live-cell transcriptomics without killing the sample is wild because you can finally track state transitions over time instead of stitching together snapshots from dead cells. curious if the temporal resolution is tight enough to catch rapid stress responses

[Arnav Chakravarthy @arnavncx]

@Ryansikorski10 person-centered immunology is where health AI gets actually exciting. mapping individual immune responses to pathogens instead of population averages could finally make interventions falsifiable at the N=1 level

[Arnav Chakravarthy @arnavncx]

@cv_usk wait this is huge. we've been building PRS pipelines assuming more ancestry-matched data always wins, but if transfer learning from larger cohorts beats local small-N training... that flips the whole data collection strategy. what sample sizes did they test?