Papers GPT

What is Papers GPT?

Papers GPT is an experimental AI-powered platform created by Jesse Zhang — technologist, founder of Lowkey (acquired by Niantic), and builder at the intersection of web development, machine learning, and quantitative finance. Unlike a traditional portfolio site, Papers GPT functions as a living lab: each project is a self-contained “paper” — not in PDF form, but as interactive, deployable code — demonstrating novel applications across real-time systems, cryptographic primitives, generative AI, and financial data engineering.

How to use Papers GPT?

Visit PapersGPT.dev (not JesseZhang.org) to access the live environment. The interface is intentionally minimal — no navigation bar, no login — just a curated grid of experiments. Click any tile to launch its full experience: run a multiplayer game in your browser, generate DALL·E images from voice input, explore balance sheets with dynamic visualizations, or verify a Mastermind solution using zero-knowledge proofs — all without installation or sign-up.

Key Features From Papers GPT

Interactive ML Demos

Real-Time Multiplayer Infrastructure

Financial Statement Engine (SEC + Yahoo Finance API)

Zero-Knowledge Proof Playground

Voice-to-Image Pipeline (Whisper + DALL·E 3)

On-Chain NFT Ownership Explorer (Ethereum L1 & L2)

Papers GPT's Use Cases

Testing cutting-edge web architectures (WebSockets, WASM, edge functions)

Running low-latency games with deterministic sync logic

Analyzing quarterly earnings through interactive, multi-dimensional charts

Learning ZK concepts via playable, transparent cryptography demos

Prototyping multimodal AI workflows — speech → text → image

Auditing NFT collection rarity, floor trends, and wallet-level holdings

FAQ from Papers GPT

What is Papers GPT?

Papers GPT is an open, executable archive of technical experiments — not a static blog or resume. It's where Jesse Zhang publishes working implementations of ideas that bridge theory and production-grade interactivity, spanning web infrastructure, ML systems design, and algorithmic finance.

How to use Papers GPT?

Go to PapersGPT.dev. Every experiment is self-hosted, client-side-first, and optimized for instant load. No backend dependencies — just click, observe, and optionally inspect the source code (all repos are public on GitHub).

Who is behind Papers GPT?

Jesse Zhang — Harvard CS alum, former Lowkey CEO (acquired by Niantic in 2022), and current engineer on Niantic’s Social Systems team. His work emphasizes *deployable insight*: turning research papers, whitepapers, and conference talks into functional, shareable web experiences.

What domains does Papers GPT cover?

Three core pillars: Web Dev (real-time sync, edge rendering, WebAssembly modules), ML (voice pipelines, model distillation, prompt engineering interfaces), and Investment Experiments (financial data normalization, valuation heuristics, on-chain alpha signals).

Are the multiplayer games production-ready?

Yes — they’re built with battle-tested stack choices: Socket.IO for matchmaking, deterministic lockstep simulation, and client-side prediction. Games like *Camel Up Live* support 10+ concurrent players with sub-100ms round-trip latency — all running entirely in-browser.

Can I analyze any public company’s financials?

Absolutely. The Financials Visualizer pulls raw 10-Q/K filings via SEC EDGAR and enriches them with market data. You can toggle between income statements, cash flow, and balance sheets — with custom time-series overlays, YoY delta heatmaps, and ratio calculators.

How does the Zero-Knowledge demo work?

It implements a simplified zk-SNARK circuit (using Circom + SnarkJS) to prove knowledge of the secret Mastermind code *without revealing it*. Users see proof generation time, verification status, and circuit size — making abstract crypto concepts tangible and measurable.

Is the voice-to-DALL·E feature fully automated?

Yes — it uses Whisper for transcription, OpenAI’s API for prompt refinement, then DALL·E 3 for image synthesis — all orchestrated via serverless functions. Each generated image includes metadata (prompt, timestamp, confidence score) and is cached for reproducibility.

Does the NFT explorer support cross-chain lookups?

Currently focused on Ethereum mainnet and Arbitrum, with support for ERC-721 and ERC-1155 standards. Future updates will add Base and Optimism. The tool highlights concentration metrics, top holders, trait distribution, and verified contract links — designed for both collectors and analysts.