{"id":"66is1jl-kII","title":"Your Data Isn't Yours — How Nova Fixes That | Julian Carbonnell","description":"Julian Carbonnell is a tech entrepreneur with a PhD in urban science who's spent years working across real estate, smart cities, governance, and machine learning. His latest project, Nova, is a decentralized file sharing SDK built on NEAR that lets users upload encrypted data once, control exactly who can access it (humans or AI agents), and revoke that access instantly.\n\nIn this episode, we get into:\n\n→ Why Julian believes the current model of data ownership is broken and what self-sovereignty actually looks like in practice\n→ How Nova uses IPFS, client-side encryption, on-chain access controls, and Shade Agents (TEEs) to create a fully secure data layer\n→ The real cost problem with centralized storage — and why paying once to upload beats paying monthly to keep your own data private\n→ A live demo of encrypted file sharing between two sessions using wallet and email login\n→ Federated learning use cases: how hospitals could train better AI models on shared patient data without exposing it to each other\n→ Julian's conversation with a local government official about replacing paper-based bureaucracy with upload-once infrastructure\n→ Why he chose NEAR (scalability, low fees, chain abstraction) and how Nova evolved organically from a music sharing app for independent artists\n→ What Estonia's digital nation taught him about building citizen-first infrastructure\n→ Where Nova goes from here: hackathon sponsorship, SDK adoption, and early business development\n\nNova SDK: https://nova-sdk.com\nNEAR Docs: https://docs.near.org\n\nTimestamps:\n0:00 — Intro & Julian's background\n1:03 — What led Julian to NEAR and Nova\n3:29 — What is Nova in Julian's own words\n5:15 — Nova vs Dropbox, S3, and centralized file sharing\n7:46 — Uploading once to IPFS — what that means for non-crypto people\n9:36 — Why decentralized storage matters for AI and ML\n10:48 — Data quality, garbage in garbage out, and immutable datasets\n12:18 — The black box trust problem with AI training\n13:11 — Companies leaking data to external AI providers\n14:41 — Federated learning: hospitals training AI without sharing patient data\n16:07 — Nova's tech stack: client-side encryption, IPFS, on-chain access, TEEs\n17:00 — How the architecture evolved organically\n19:11 — Shade Agents, off-chain key storage, and bulletproof security\n21:07 — Practical design vs opinionated design\n22:23 — Why TEEs? Using the latest NEAR tech\n23:43 — The real threat model: self-sovereignty over fear of leaks\n24:36 — Why NEAR as the blockchain\n27:50 — What \"AI ready\" actually means for Nova\n29:18 — Nova as an SDK — the vision for embedding it everywhere\n30:36 — Not all data belongs on decentralized storage\n31:15 — Local government use case: replacing paper administration\n33:23 — The recurring cost trap: privacy tied to monthly subscriptions\n35:41 — Adoption strategy and what people might build with Nova\n37:26 — Origin story: Nova was born from a music sharing app for artists\n39:38 — Nova as agnostic infrastructure\n40:28 — What Nova's success would signal beyond business\n42:04 — Data privacy regulation, GDPR, and how Nova fits in\n45:27 — Inspired by an academic paper from Taiwan on blockchain privacy\n46:55 — Estonia's digital nation and citizen-first design\n50:45 — Live demo: encrypted file sharing with Nova\n56:24 — Mass adoption, non-tech users, and the pay-once model\n58:19 — What's next for Nova and Julian\n\n#NEAR #Web3 #DataPrivacy #DecentralizedStorage #AI #SelfSovereignty #IPFS #Nova","channel_name":"NEAR DevHub","published_at":"2026-03-26T00:00:00.000Z","thumbnail_url":"https://img.youtube.com/vi/66is1jl-kII/hqdefault.jpg","transcript_text":null,"transcript_synced_at":"2026-11-03T03:48:33.153Z","segments":[],"youtube_url":"https://youtube.com/watch?v=66is1jl-kII","url":"https://near-pedia.xyz/videos/66is1jl-kII"}