An annotated mixtape of 2021’s AI breakthroughs

What it does

This repository is a curated reading list of 38 influential AI papers released in 2021. Maintainer Louis Bouchard compiles each entry with a short video walkthrough, a plain-English article, and links to the original paper and code repository when available. It is essentially editorial infrastructure: the repo itself contains no runnable software, but it saves you from trawling arXiv and YouTube to reconstruct the year’s research narrative.

The interesting bit

The list doubles as a syllabus for the generative-AI inflection point, capturing DALL·E, early NeRF variants, StyleGAN manipulations, and diffusion-adjacent work while they were still fresh. Because Bouchard produced the videos and articles himself, the tone stays consistent; you are not bouncing between unrelated conference talks and hastily written summaries.

Key highlights

• 38 papers, each with a dedicated ~5-minute video and companion blog post.
• Coverage spans text-to-image synthesis, 3D reconstruction, video generation, and on-device privacy-preserving ML.
• A 15-minute “2021 rewind” video stitches the highlights into a single narrative.
• Sister repository isolates the top 10 computer-vision papers for narrower focus.
• Explicitly notes when code is unavailable, so you know whether a paper is reproducible before you click.

Caveats

• The repository is strictly a directory of external links; there is no code to run or clone here.
• Selection favors generative and vision research, so the list is not a balanced survey of all AI subfields.

Verdict

Worth bookmarking if you are a practitioner or student trying to map the 2021 research landscape without reading three dozen papers raw. Skip it if you are hunting for a unified codebase or training pipeline to run out of the box.