The original GPT-3 demo reel, still curated

What it does — This is a classic “awesome list”: a single markdown file serving as a hand-curated directory of early GPT-3 demos, articles, GitHub repos, and commercial products. It catalogues Twitter threads, blog posts, and tools from the months after the OpenAI API launched, grouping them by task—program generation, text completion, search, and even game creation.

The interesting bit — The list captures the initial shock of developers realizing a language model could generate React components from variable names, translate legal jargon, or attempt chess. Reading it now feels like browsing a fossil record of prompt-engineering before ChatGPT made those tricks commonplace.

Key highlights —

• Demos span code generation (SQL, Python, LaTeX), design tools (Figma plugins, layout generators), and text tasks (email drafting, poetry, legal simplification)
• Includes a reading list with technical overviews, hype-tempering essays, and Gwern’s creative-fiction experiments
• Links to early tooling like the GPT-3 Sandbox and a few ChatGPT wrappers
• Serves as a snapshot of pre-ChatGPT application brainstorming, when raw API access felt novel

Caveats —

• This is strictly a link directory, not a runnable project; there is no code to execute beyond the README itself
• The focus is overwhelmingly on the original GPT-3 API wave, so modern API features and later models are largely missing

Verdict — Historians of AI hype and researchers tracing the evolution of LLM applications will find it a useful index; builders looking for current tooling should treat it as a museum, not a manual.