@pleometric
Account based inSouth America
About this account
- Account based in
- South America
- Connected via
- South America App Store
Account-level information from X, not a live location or the device used for a specific post.
Purple Penguin
The Pleorama
Joined February 2024
- Tweets694
- Following554
- Followers9.4K
- Likes22.6K
I like how meta can scale into anything they want to do, regardless of product quality
$META’S MUSE IS OUTPACING CHATGPT’S EARLY MOBILE LAUNCH ON SEVERAL METRICS
Muse reached 2.8M downloads in its first 12 days, across both app stores.
On a more comparable U.S. + Canada iOS basis:
• Muse: 1.8M downloads in first 12 days
• ChatGPT: 1.3M
Muse also set a new U.S. daily download record of 264K on Sept. 19, its 3rd straight day above 200K.
ChatGPT didn’t cross 200K U.S. daily downloads until roughly a year after launch.
Engagement is also running ahead:
- Muse reached 448K U.S. DAUs by day 10
- ChatGPT took 49 days to reach roughly 450K U.S. DAUs
- By day 12, Muse was at 642K U.S. mobile DAUs vs. 231K for ChatGPT at the same point
- Even restricting Muse to U.S. iOS, DAUs were 359K
Source: Sensor Tower
I wrote a brief guide on how to make brainrot. I'm trying to compress most of the questions into a single article.
I'll be trying to publish something every day over the coming week. If you have questions you want explored let me know.
pleometric.net/articles/brie…
I explored, tested and validated every single pipeline I have with Astra, it has thoroughly one shotted me. It could do any audiovisual task I threw at it with varying degrees of success but at its WORST it's a very competent intern fully able to course correct... one shotted...
Peep those digits...
If you frame the joke not on the video but around it, you too can enjoy AI without the negative externalities getting in the way of the joke!
Naively I would think so sort of state object, like a scene description json, would be better than checking raw pixel data. Then I remember the bitter lesson...
maybe compressing the images until I reach some arbitrary threshold for checking state?
surely this exists...
while Astra has impressive applications for video editing, it is still absurdly expensive to run.
Most token spend comes from visual state checks. Even native MCP implementations like Resolve only mitigate the issue slightly. No program has the correct shape for it... I have found that outsourcing all visual tasks to gemini cheapens the process, but it's very tool call expensive.
You know what's the most expensive Astra editing operation? Audio! Beat matching, spectral analysis etc. Very, very, very expensive.
while Astra has impressive applications for video editing, it is still absurdly expensive to run.
Most token spend comes from visual state checks. Even native MCP implementations like Resolve only mitigate the issue slightly. No program has the correct shape for it... I have found that outsourcing all visual tasks to gemini cheapens the process, but it's very tool call expensive.
while Astra has impressive applications for video editing, it is still absurdly expensive to run.
Most token spend comes from visual state checks. Even native MCP implementations like Resolve only mitigate the issue slightly. No program has the correct shape for it... I have found that outsourcing all visual tasks to gemini cheapens the process, but it's very tool call expensive.
