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.