Source

Developer data management for an edge-first world

[Source @EdgeOfSource]

@isocroft @getifyX Merkle-CRDT is a whole other animal

[Source @EdgeOfSource]

DefraDB v1.0 RC1 is here. Our first stable release candidate after 25 pre-production releases.
A peer-to-peer document database for edge AI and offline-first applications. Data syncs across devices using Merkle CRDTs. No central server required.
https://t.co/wYoujGHpOG

[Source @EdgeOfSource]

Sourcerer 🧙‍♂️ @getifyX spitting facts about passkeys @nordicjs

[Source @EdgeOfSource]

@joodalooped @getifyX @zero__ms @convex That exists because people didn't realize that we had explanations for each of the edge-first statements, and they would ask why it had to be p2p, verifiable, etc. So 2 seconds of scrolling vs constantly getting asked the same questions, even if the answers were on the site

[Reece Williams @Reecepbcups_]

I’m hiring backend / protocol engineers to build a new product in the Penumbra codebase

DMs are open if you understand these systems like this / namada

[Source @EdgeOfSource]

📚 System Design: Proxy Re-Encryption

**Problem:** Share encrypted data without exposing content
**Solution:** Proxy Re-Encryption

- Alice encrypts data with her key
- Proxy converts to Bob's key format
- Bob decrypts with his key
- Proxy never sees plaintext

Used in secure sharing systems like Source Orbis 🔄

[Source @EdgeOfSource]

🌐 Tech Philosophy: Edge-First Computing

- **Cloud-First**: Send data to distant servers
- **Edge-First**: Process data where it's created

**Advantages**:
- Lower latency (faster responses)
- Better privacy (data stays local)
- Offline capability
- Reduced bandwidth costs

The shift from centralized to distributed computing 🚀

[Source @EdgeOfSource]

💡 Concept Explained: Threshold Cryptography

Instead of "1 key = 1 signature," threshold crypto uses an M-of-N scheme:

- Split key into N shares
- Any M shares can create a signature
- Individual shares are useless alone

This increases security and fault tolerance.

Example: A 3-of-5 system survives 2 node failures. ⚖️

[Source @EdgeOfSource]

🔬 Algorithm Spotlight: Merkle DAGs

- DAG = Directed Acyclic Graph (like a family tree, no loops)

**Merkle DAG benefits:**

- Each node has a cryptographic hash
- Changes create new hashes (tamper evidence)
- Efficient verification of large datasets
- Used in Git, IPFS, and DefraDB

Think "cryptographic data structures that verify themselves" 🔗

[Source @EdgeOfSource]

📖 Tech Tutorial: Distributed Key Generation (DKG)

Traditional crypto: 1 private key = 1 point of failure

DKG approach:
- Multiple parties collaborate
- Generate key shares (no one has full key)
- Threshold operations (e.g., 3 of 5 needed)
- Reconstruct signatures collectively

Result: Security through distribution, not trust 🛡️

[Source @EdgeOfSource]

🧠 Deep Dive: Zero Knowledge Proofs

ZK Proof = Prove you know something without revealing what you know

Real examples:
- 🔐 Prove you're over 18 without showing your birthdate
- 🔐 Prove you have funds without revealing your balance
- 🔐 Prove identity without exposing personal data

3 properties: Completeness, Soundness, Zero-Knowledge 🎯

[Source @EdgeOfSource]

🎓 Learning Corner: Content-Addressable Data

Traditional web: "Where is the data?" (URLs point to locations)

Content-addressable: "What is the data?" (Hash identifies content)

Benefits:
- ✅ Data integrity verification
- ✅ Deduplication
- ✅ Decentralized storage
- ✅ No broken links

[Source @EdgeOfSource]

🧠 @GoogleDevs' Gemma 3 270M shows the future: specialized #AI models for specific tasks.

While they're solving deployment at scale, Source is solving the data layer:

✅ Edge-first data management for local AI
✅ Decentralized model versioning with DefraDB
✅ Privacy-preserving inference with ZK protocols

Small models + edge compute = true AI democratization 🚀

https://t.co/7pWxVIBQcd

[Source @EdgeOfSource]

📚 Tech Education: What are CRDTs?

CRDT = Conflict-free Replicated Data Type

Think Google Docs collaboration, but for databases:

- Multiple users edit simultaneously
- No "merge conflicts" or data loss
- Mathematical guarantees of consistency
- Works offline, syncs when connected

DefraDB uses Merkle CRDTs for added verification 🔍

[Source @EdgeOfSource]

My database is so eventually consistent that by the time it converges, I'll have already retired and started a goat farm 🐐

[Source @EdgeOfSource]

Zero Knowledge Proofs: "I can prove I know something without telling you what it is!"

My imposter syndrome: "Finally, a cryptographic protocol that gets me." 🤐

[Source @EdgeOfSource]

My code is like a fine wine: it gets better with age, but mostly, it just gives me headaches. 🍷🤕

[Source @EdgeOfSource]

My distributed system is so distributed that I need a distributed system to find my distributed system. 🤔

[Source @EdgeOfSource]

Building collaborative features shouldn't require a PhD in distributed systems. DefraDB makes it simple:

✅ Real-time collaboration built-in
✅ Offline editing with perfect sync
✅ No conflict resolution logic needed
✅ Works with any data structure

Explore real examples and documentation: https://t.co/2TXPHIyhd2

[Source @EdgeOfSource]

5/ The magic: It works offline too. Make changes locally, sync later, never lose work.

[Source @EdgeOfSource]

Why collaborative software is so complex to build:

[Source @EdgeOfSource]

4/ Think Google Docs, but for any data structure: lists, maps, counters, text, anything.