December 30, 2025

Cartoons vs Code: Grab the popcorn

Animated AI

YouTube AI cartoons spark praise, side‑eyes, and a code‑or‑die backlash

TLDR: An indie creator dropped MIT-licensed animated explainers for neural networks on YouTube/Patreon. The crowd split: fans say visuals finally make it click, skeptics say “just build one,” and veterans point to classic resources while asking for transformer explainers next—highlighting the ongoing fight over how we learn AI.

Animated explainers just dropped for the brainy world of neural networks, and the internet immediately split into camps. The creator walks viewers through “convolution” (a tiny window sliding over a picture), “padding” (adding borders), “stride” (how big the steps are), and even flashy “pixel shuffle,” all open under MIT. Fans are cheering that pictures beat equations, with one calling it a “fantastic educational resource,” while another swears calculus-style visuals of bumpy hills and downhill “gradient” moves finally click. Requests flood in: “Do Transformers and Attention next!”

But then the backlash struts in. One skeptic shrugs, “I don’t think these are useful at all,” insisting you learn more by coding a tiny network or blurring an image yourself. Cue the Cartoons vs Code showdown. Meanwhile, the old guard rolls up with receipts: classics like the legendary conv_arithmetic GIFs get name‑checked, and a proud alum drops their own CNN cheatsheet. The memes write themselves—someone joked that “Same vs Valid” sounds like pizza sizes and another that the thread has “extra padding.” The vibe: great on‑ramp for newbies, not a magic pill for mastery, and everyone’s chanting for Season 2 starring Transformers

Key Points

  • Animated instructional resources focus on convolutional operations in neural networks.
  • Padding modes are covered, contrasting Valid (no padding) with Same ([1,1,1,1] padding).
  • Stride configurations (1 and 2) and their combinations with padding are demonstrated.
  • Grouped, depthwise, and depthwise-separable convolutions are illustrated with examples (1, 2, and 8 groups).
  • Pixel shuffle and unshuffle techniques for 2x2 and 3x3 block sizes are shown; materials are MIT-licensed.

Hottest takes

"This is a fantastic educational resource!" — amkharg26
"Nice! I made my own version of this many years ago" — jerpint
"I don't think these are useful at all." — throwaway2027
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.