Gemini 4 Argon does look very different in the Gemini app and Antigravity. But this 7-second video ate 10% of my weekly quota.
I don’t think Pro users can build anything serious in Antigravity at this rate. 🥺
I'm a $20/month peasant. I can't afford to run real benchmarks.
So ignore the numbers. Trust the vibes. That's all we carbon-based users have.
Hands up if you've felt this.
@thsottiaux @OfficialLoganK
ChatGPT didn’t just make me faster at work. It expanded what I could build: a mechanical optimization and failure-analysis tool, a private AI workspace, and automations I can run from halfway around the world. That’s not a chatbot. That’s leverage. @OpenAI #ChatGPTAtWork
My annual ChatGPT plan was about to expire, and I was considering switching to monthly so I could spend more time testing competitors. Then GPT-5.6 Sol arrived. After testing it on a real mechanical-engineering workflow, I renewed for another full year.
Before GPT-5.6, Fable 5 was my default model in my company’s GitHub Copilot account. I’m now shifting more work to Sol because the output quality is comparable, the engineering reasoning fits my workflow better, and the value is stronger.
On my personal account, Codex is my only choice，helped me build a multi-agent platform where agents can post, discuss projects, and hold meetings. I’m also creating personal tools and an interactive haunted-hotel atlas. Codex makes my ideas move faster. @thsottiaux
GPT-5.6 Sol found the bugs, then Codex lost the features while fixing them.
@thsottiaux @sama
For engineering tools, “fix only the identified issues and preserve all existing behavior” has to be reliable. A clear before/after feature inventory, regression tests, and an easy rollback checkpoint would have saved a lot of time and quota here.