Skip to content

Instantly share code, notes, and snippets.

Human-Archive

Egocentric hand-tracking annotation platform with difficulty-aware frame classification.

Built an egocentric hand-tracking annotation platform. Core challenge: automatically classify first-person video frames into 5 categories — no_hands, low_lighting, occluded, dexterous_pose, easy.

Approach 1 (Naive baseline)

"Started simple: uniform 2fps sampling + single MediaPipe detector in IMAGE mode. Got 70% no_hands, 20% dexterous, 10% easy — but zero occlusion or low-lighting detection. Clear failure: occlusion, lighting, and complex poses were invisible."

sequenceDiagram
    autonumber
    
    participant User as User/dApp
    participant RPC as Validator RPC<br/>:8899
    participant Queue as TX Queue<br/>Vec<Transaction>
    participant Policy as Batch Policy<br/>Hybrid/Adaptive/Auto
    participant Executor as Deterministic Executor<br/>BTreeMap<String,u64>

Obsidian Sync Engine

1 · Full Architecture

sequenceDiagram
    autonumber
    participant VR  as Vault Repos
    participant Dev as Developer
    participant GH  as GitHub CI

To understand project and architecture, please see this README - https://github.com/MdSadiqMd/TraceZero

  1. How to solve small user base

    • If 3 users buy credits at 14:00, 14:05, 14:10, and 3 deposits arrive via Tor at 14:30, 14:35, 14:40, the relayer has a strong timing correlation. With a small user base it can be easily trackable
    • And when there are two users in the pool it means there is 50% of chance of tracing the person, narrowing it down, how many users in the anonymity set is safe, and how to make sure the first user is not exposed ?
  2. The deposit wallet needs SOL to fund pool deposits. Where does that SOL come from?, currently relayer operator periodically transfers funds from treasury → deposit wallet (off-chain), but this transfer is on-chain. How is the deposit wallet funded without creating an on-chain link to the treasury?

  3. The client generates a new ephemeral X25519 key per deposit request. But the relayer's ECDH key is a StaticSecret generated once at startup and reused for ALL

Incognitus — Frontend Specification

Project: Incognitus — On-Chain Central Limit Order Book (CLOB) Chain: Solana Scope: Frontend client — WASM + WebGPU + React Hackathon: Colosseum Frontier (April 6 – May 11, 2026)

incognitus-frontend-spec.md 27 KB Nomadic° — 8:08 PM

Folder Structure

src/
├── types/
│   └── user.ts
├── hooks/
│   └── useCreateUser.ts
├── components/
│   └── CreateUserForm.tsx

ProveKit WASM Demo - Complete Architecture Documentation

System Overview

The ProveKit WASM Demo is a browser-based zero-knowledge proof system with GPU acceleration. It consists of three main applications and a comprehensive build/verification pipeline.


1. User Flow Diagram

Incognitus — User Flow Sequence Diagrams

Every user-facing flow in the system, traced end-to-end with full detail. Covers onboarding, deposits, limit orders, market orders, matching, settlement, cancellation, withdrawal, and the crank lifecycle. Each diagram shows every account touched, every validation performed, and every state change made.


Table of Contents

Prompt Log

Prompt 1

Date: 2026-03-24 | Time: Session

Prompt: "Read docs/documentation.md to understand this project more, and from now write all prompts in docs/prompts.md with short summaries as a log."

Summary: Set up docs/prompts.md as a running log file. Read docs/documentation.md which contains comprehensive ZK proof documentation for ProveKit — 20 mathematical topics, 100+ interview questions, flowcharts, comparison tables, security considerations, and a glossary. Ready for next tasks.


ProveKit Documentation

What does this project do?

ProveKit is a zero-knowledge proof toolkit optimized for mobile devices by the World Foundation.

ProveKit compiles Noir circuits (a domain-specific language for zero-knowledge proofs) into R1CS (Rank-1 Constraint System) constraints and generates/verifies WHIR proofs. The workflow is:

  1. Takes Noir circuits (written in the Noir language) as input
  2. Compiles them to R1CS constraint matrices using an optimized compiler