New data: ChatGPT uninstalls surged by 295% in a single weekend. Trust in OpenAI collapsed fast. That’s the central story from a creator who covers every significant AI development in the space, and the numbers behind it are worth unpacking.
Here’s what triggered it
The original poster breaks down an explosive chain of events that reshaped the AI landscape this week. Anthropic drew two firm red lines with the Pentagon: no surveillance of US citizens, no fully autonomous weapons. The Department of War rejected those terms and designated Anthropic a supply chain risk. OpenAI stepped in the same day, signed a contract, and publicly claimed the same red lines. The public didn’t buy it. Within days, Claude shot to the number one most downloaded app in the App Store.
And the business numbers back it up. Anthropic is nearing a $20 billion annual revenue run rate, more than doubling from last year. A Ramp spending chart shows corporate AI spend flipping from overwhelmingly OpenAI to now favoring Anthropic. The trust gap is real, and it’s showing up in the data.
This week’s model releases
- 🔹 GPT-5.4 is out for paid ChatGPT users (Plus, Team, Pro). It adds native computer use, a 1 million token context window, and a new “tool search” feature that reduces token bloat when models call external tools. The creator notes it’s most impactful for coders and API users. Everyday chatters will notice less.
- 🔹 NotebookLM now generates cinematic video overviews powered by VO3 animations, not just static slideshows. The creator tested it on a conspiracy theory notebook and the output looked like professional motion graphics. The catch: it requires the $250/month Ultra plan for now.
- 🔹 Google Canvas launched inside AI Mode, bringing a code-and-preview workspace directly into Google Search, similar to what you’d find in Claude or ChatGPT.
Things to watch
Meta’s AI smart glasses are facing lawsuits in the US and regulatory scrutiny in the UK after reports surfaced that human annotators were reviewing sensitive footage from users whose privacy settings weren’t configured correctly. The creator flags that sharing footage with Meta appeared to be on by default unless manually disabled.
Also worth noting: a new anti-recording device called Spectre 1 promises to jam nearby audio recorders. But the expert raises a fair question about whether it disrupts legitimate devices like phones and earbuds. At roughly $1,000, it’s more concept than practical product right now.
The takeaway
The takeaway from this week is that model specs matter less than trust. Benchmarks moved. Red lines moved. But what actually moved users was a company holding its position under pressure. The creator suspects the supply chain risk designation against Anthropic won’t stick long-term and that some kind of compromise with the Pentagon is the likely outcome, though he admits he’s been surprised before.
Watch the full video for the complete breakdown of every model drop, the OpenAI staff fallout, and the Anthropic internal memo leak.