@raythurnvoid
Account based inPortugal!
About this account
- Account based in
- Portugal
- Connected via
- Web
X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
Fullstack developer @ Sybill
Joined September 2015
- Tweets1.7K
- Following520
- Followers65
- Likes3.4K
I asked Grok 4.7 to do some task on the browser and it had issues. It decided to create its own custom computer use using PowerShell
OpenAI going through a compute crisis like Anthropic a few months ago
🚨 OPENAI Has a big problem
OpenAI appears to have quietly removed the 5× and 20× usage labels from its Pro plans
Previously:
> $100 Pro = 5× more usage than Plus
> $200 Pro = 20× more usage than Plus
Now both simply say: “More usage than Plus”
OpenAI’s own help page still describes the tiers as 5× and 20×, so the UI change is especially interesting
Why is OpenAI so bad at being transparent It’s not the first time they’ve done something like this??
I downloaded 5 seconds from a 1hr video from YT
1 year ago this was a 30 minutes process to find a yt downloader that was not scamming, find a video editor without watermarks, open the video, trim and export the video.
Now I asked grok 4.7 to do it, it did in 2 minutes in bash
Opus 5.5 is as far ahead to Astra and Fable 5.1 as Fable 5 was far ahead to GPT 5.5 and Opus 4.8
The level of depth this model has is incredible, it understands the world far better than anything before
If AI stops here, I'm more than happy
6 Months ago I would have never believed this technology would get this good.
So today I edited a video and I've been doing it with AI
The last time I did video editing was like 10 years ago with Sony Vegas, now I'm using remotion with AI
I wonder where this will go:
Ray Thurn Void retweeted
GPT 6 Astra made this video entirely with code.
Same prompt that I used for the Claude Opus 5.5 video.
Anthropic is so far ahead right now.
GPT models' tendency to overengineer things has finally (and sadly) cost me money.
It wrote 150M rows to a database just for "capacity testing" and later apologized for it.
GPT-6 Astra + GPT-6 Sol
Not good.
ALT Spend Summary dashboard: Total cost is $102.90 for September 9–24, 2026, with a projected 30-day cycle cost of $192.94 and an average daily cost of $6.43. A donut chart shows $102.00 for D1 rows written and $0.90 for R2 data storage.