Try AIQORA:
Start free and get 10 extra credits with code MATTVIDPRO.
This segment is sponsored by AIQORA, an AI video platform from Germany.
I was out sick for five days and the AI world did not wait. Gemini 4 Argon is out, and it’s genuinely the first Gemini that feels like a big model. In one same-prompt game build it cost 43 cents to Opus 5.5’s $2.43. So what’s the catch? I put Argon through a real Roblox build in Google’s Antigravity until it ran out of quota. I also cover Meta’s Muse Wonder and the open-weight Muse Spark Open, an OpenAI employee arrested over a hidden recording device, a public OpenAI model helping break Chrome for a $250K bounty, and why I just upgraded my Claude plan anyway.
▼ Links from today’s video:
Google: Introducing Gemini 4 Argon
VentureBeat: Argon broad availability and intro API pricing
Meta: Muse Wonder model card
Meta: Muse Spark Open docs
Roblox Studio MCP server (used for the Antigravity build)
CNBC: OpenAI employee arrested over hidden recording device
TechCrunch: FBI arrests OpenAI employee accused of planting a recording device
The Hacker News: Researchers used a public OpenAI model to build a Chrome exploit
Anthropic Engineering: Redesigning source control at Anthropic
Axios: OpenAI DevDay 2026
MattVidPro Discord:
Thanks to Asmir (Argon vs Opus 5.5 pigeon build), Universal Modder (Argon vs Muse Wonder game comparison), Kicking Keys, ZCW and Dave Shapiro, whose posts I showed in the video. The cold open was made with Seedance 2.5.
Clarifications: 43¢ vs $2.43 is the API cost of one same-prompt build from Asmir’s test, not a general price ratio. Argon’s $2/$10 per 1M token pricing is a limited-time intro rate. The recording-device case is an allegation and the person has not been convicted.
0:00 I got rebuilt (Seedance cold open)
0:18 Sick for 5 days. Here’s what I missed
0:39 Gemini 4 Argon finally feels like a big model
1:23 Argon vs Meta’s Muse Wonder
2:23 Muse Wonder pricing: the harsh reality
3:01 Muse Spark Open: huge open weights
3:43 Same prompt: Argon 43¢ vs Opus 5.5 $2.43
5:06 Benchmarks feel useless
5:35 My Roblox build in Antigravity
6:53 Out of quota
7:36 The benchmarks don’t add up
9:27 Hidden recording device at OpenAI
10:59 A public OpenAI model helped break Chrome
11:55 An AI anime recipe
12:51 AIQORA (sponsor)
14:30 Why Opus 5.5 feels human
15:57 Anthropic now merges 30,000 changes a day
16:27 Outro
source
