This week on the channel +8M views this week

Fireship puts AI through horse tests and bubble doubts

Two uploads brought a new model review and an industry-sized question, while the channel collected roughly eight million weekly views.

Channel:

Fireship Story of October 31, 2026Trari Newsroom

Why this story? The channel's weekly summary: what went up this week, what is drawing views now, and where the channel stands among similar channels.

China just dropped a new frontier model... and nobody can explain how

Things we can learn from the video

Press to jump

0:00The anonymous Chatbot Arena model Kibitz is revealed as Moonshot AI's Kimi K3.1 Pro.
2:09Coding benchmark results place Kimi close to closed models such as Opus.
2:46Delaney describes Moonshot's removal of cybersecurity exploit data to reduce the model's hacking abilities.
3:46ThoroughBench tests whether models can generate interactive 3D horses using Three.js.
4:23Kimi generates a detailed horse, but its movement relies on simple mesh warping.
5:29A sponsored segment demonstrates a design tool's integration with coding agents.

Fireship gave AI two very different examinations: could a new model build a convincing animated horse, and was the business surrounding the technology in trouble?

The model test drew the week's largest new audience. "China just dropped a new frontier model... and nobody can explain how" has passed two million views, with Jeff Delaney taking Moonshot AI's Kimi K3.1 Pro from coding benchmarks to his rather more memorable ThoroughBench. The assignment: generate an interactive 3D horse using Three.js. Kimi produced a detailed animal, then let it down with crude animation.

The comments have turned that test into a running institution. Viewers are asking for a universal horse benchmark and joking that the equine experiments are their real reason for keeping up with AI developments. The episode ends with a sponsored segment.

"Did the AI bubble just pop?" arrived later and now draws the most daily views of any Fireship upload. Both new releases had similarly strong opening days, but the bubble question currently has considerably more momentum. Last week's "The last real programmers are dying out..." also remains active.

Across the channel, daily viewing is a little below its usual level, even as subscribers grew over the week. Among nearby tech channels, Low Level and Matt Wolfe are having particularly elevated days. Fireship nevertheless recorded the largest weekly view gain in this peer group, with a pair of uploads that moved from a badly walking horse to AI's financial footing.

Published on Trari

Where next?

Also in Technology companies and softwareViews jump 8.9x the channel's usual paceMatt Wolfe puts his newsletter strategy before an AI panelNate Herk's "Roast Room" workflow gives the channel's automation coverage a practical test as daily views rise sharply.Channel: Matt Wolfe · Story of October 31, 2026

A different discovery todayVideo taking off 47x its usual first dayDiane Sawyer's Patrick Stewart interview closes World News TonightThe October 29 edition also covers Alaska's Senate probes and a Pentagon firing, but its comment section belongs largely to the actor.Channel: ABC News · Story of October 31, 2026

What viewers say

I only watch fireship to keep up with the development of horse tinder
@allrevenuegoneSee it on YouTube
AI companies: trillion parameters.

Me: show me the horse.
@Davidayo_See it on YouTube
We need an oficial horse benchmark score for all AIs
@forealoveSee it on YouTube

3 more comments

Well you forgot to add "make no mistakes". No wonder the horse moves like that.
@MrbiernotSee it on YouTube
Horse benchmark should be used as universal benchmark for all new models
@GenMathArtSee it on YouTube
Last time i was this early AI was fancy autocomplete
@fff-kd8kgSee it on YouTube

Public comments from the video's page on YouTube