Show HN: Jmail – Google Suite for Epstein files

Original Article ↗ Hacker News Discussion ↗

Overall reception & UX

  • Many commenters praise the project as “next level,” both for the concept (a Gmail-style interface on the Epstein corpus) and for the speed/quality of execution.
  • People especially like the intuitive browsing, search, flight visualizer, and Gmail-like touches (e.g. jump-to-beginning/end buttons that Gmail lacks).
  • There are some UI/UX bug reports (back button issues, limited image navigation, mobile quirks) and feature requests (prev/next for emails/photos, better attachment views).

Technical implementation

  • The suite now includes Jmail, JPhotos, JDrive, JAmazon, JFlights, and “Jemini,” an LLM assistant over the files.
  • Data is parsed from PDFs and images using LLM-based tools (e.g. reducto), multimodal classification, and vector embeddings (turbopuffer); text messages and more structured UIs are planned.
  • Infrastructure: Next.js on Vercel, Cloudflare R2 for assets (free egress, heavy caching), Neon for the database. Earlier versions ran on Railway + Cloudflare cache and reportedly handled millions of visitors at very low cost.
  • One contributor is experimenting with Apple’s SHARP model to generate Gaussian splats from single images for 3D-style visualizations.

Data sources & completeness

  • Jmail currently aggregates three datasets:
    • DOJ “Epstein Files” release
    • House Oversight Committee drops (including large November dump, flight logs, photos)
    • Yahoo emails originally from DDoSecrets, manually redacted by a partner organization.
  • Commenters note the DOJ release is a tiny fraction of the reported archive size and heavily redacted; others highlight that Oversight and Yahoo sets contain lots of non-overlapping material.

Redactions, authenticity, and alleged tampering

  • Several comments assert over-redaction beyond what the law allows and characterize DOJ behavior as “malicious compliance.”
  • There’s debate over claims that some images were “planted” or misleadingly framed (e.g. a redacted photo involving well-known figures and children). Others push back, suggesting procedural redaction of all minors or ordinary evidence handling; overall, motives and provenance are labeled as unclear.
  • People discuss how redaction might work (automation, fragmented review, legal constraints, reporting of redactions to Congress).

Ethics, privacy, and justice

  • Some are uncomfortable with turning a victim-heavy case into a slick, quasi-parodic product suite and with the voyeurism of combing through private digital life.
  • Others argue victims themselves have pushed for full release to force accountability, and that public scrutiny is necessary when institutions are suspected of protecting powerful people.
  • There’s skepticism that disclosures will seriously harm major political figures or change systemic impunity; some see this as potential distraction or “reality show,” others as a necessary step toward RICO-style or financial investigations.

Exploring the contents

  • Users find the Amazon orders, book purchases, flights, and social graph fascinating in a mundane-but-revealing way: everything from Crocs and shower heads to philosophy texts, AI/intelligence discussions, and elite-network interactions.
  • Several note how unsettling it is to see an entire two-decade digital footprint laid out, yet also how powerful structured search and LLM tools are for navigating an otherwise messy corpus.

“Cloned Gmail” and software moats

  • There’s a mini-thread pushing back on claims that the team “cloned a whole suite of products”: commenters stress this is a front-end/UI imitation over a static corpus, not a full email service with all of Gmail’s backend complexity.
  • This leads to broader discussion about how much engineering effort underlies real webmail (spam filtering, scale, protocols) and whether quick UI clones actually threaten large software moats.