Comments URL: https://news.ycombinator.com/item?id=47586491

Points: 1

# Comments: 1

~40M people have DNA chip data sitting unused. Most consumer reports give you 30-300 health traits. We score 1,261+ polygenic risk scores across every major disease category, scan 400,000+ variants against ClinVar for pathogenic findings,run pharmacogenomics on 34 genes (110K+ diplotypes via CPIC), and generate an AI-powered longevity protocol.

  How the scoring works:

  - Effect allele dosages x published PGS Catalog weights
  - Percentiled against empirical distributions from 2,504 1000 Genomes Phase 3 samples across 5 ancestry groups (EUR, AFR,EAS, SAS, AMR)
  - Ancestry-aware ML correction via Ridge regression (8 features including ancestry proportions)
  - Optional deep imputation (Beagle 5.5) expands 609K chip variants to ~28M
  - Each score shows confidence level, variant coverage, and population calibration

  What you get:

  - 1,261+ polygenic risk scores organized by category (cardiovascular, neurological, metabolic, cancer, autoimmune, etc.)
  - ClinVar pathogenic variant matches (400K+ scanned)
  - Pharmacogenomics: star allele calling for 34 genes with CPIC dosing recommendations
  - Convergence detection: flags when multiple PRS + monogenic variants point at the same disease pathway
  - AI-generated longevity protocol (supplements, training, diet — all matched to your genetic profile)
  - Interactive genome explorer + 4 downloadable PDFs (Health Summary, Doctor's Report, Full Analysis, Health Protocol)
  - JSON export to feed your genetic profile to AI assistants

  Privacy (this was the main design constraint):

  Zero data retention. No accounts, no cookies, no tracking. Your DNA file is deleted from disk immediately after analysis (~2 minutes). All remaining data (reports, scores, PDFs) is purged within 2 hours. At deletion time, you receive an automated Data Deletion Certificate — a PDF with a SHA-256 attestation hash (SHA256(certificateId:jobId:deletedAt:salt)) proving when every piece of your data was destroyed. Every deletion is logged in an append-only audit trail containing zero genetic data.
   We built this specifically because 23andMe's bankruptcy exposed 15 million people's DNA. There is literally nothing to breach here.

  Honest caveats:

  - PRS capture common variant risk only — they're not crystal balls, they show relative genetic risk vs. the population
  - Most PGS models were developed in European cohorts; cross-ancestry performance varies
  - This is a research/educational tool, not a certified diagnostic
  - Still some rough edges — we've only tested with a handful of people so far

  We've had some surprisingly accurate results — one report independently flagged genetic variants consistent with a known medical condition that hadn't been mentioned to us.

  Giving away 5 free beta codes to the first commenters. Would genuinely love honest feedback, bug reports, and criticism.
  

Comments URL: https://news.ycombinator.com/item?id=47176628

Points: 1

# Comments: 1

Your writing style can theoretically be masked with an LLM. Your genome can't. And it doesn't just identify you -- it identifies your relatives, your disease risks, your ancestry, things you might not even know about yourself yet. The deanonymization vector here is permanent and irrevocable in a way that no amount of OPSEC can fix after the fact.

The semantic approach in this paper (interests, clues, behavioral patterns) is scary enough. Now imagine combining that with leaked genetic data. You don't even need to match writing styles when you can match someone's 23andMe profile to their health subreddit posts about conditions they're genetically predisposed to.

The structural problem is that once you've taken billions in VC, safety becomes a negotiable constraint rather than a core value. The board's fiduciary duty runs toward returns, not toward whatever was in the mission statement. PBC status doesn't change that in practice — there's basically zero enforcement mechanism.

What's wild is how fast the cycle has compressed. Google took maybe 15 years to go from "don't be evil" to removing it from the code of conduct. OpenAI took about 5 years from nonprofit to capped-profit to whatever they are now. Anthropic is speedrunning it in under 3. At this rate the next AI startup will launch as a PBC and pivot before their Series B closes.