Techno-optimist fascinated by the unconventional and psychology. Exploring new learning methods, lost in poetry, and gazing at the starry sky.
earth
Joined April 2022
- Tweets1.2K
- Following291
- Followers64
- Likes4.9K
Nova retweeted
Google says Gemini has caught up with the best models. Fair enough.
But @GeminiApp, could you benchmark the web experience against ChatGPT too?
Even in Google's own Chrome, long conversations make Gemini Web hang. ChatGPT gets slow too, but at least it stays responsive.
The models caught up. The product hasn't.
cc @antigravity
I just used a Claude banked reset, and… it doesn't change my main reset date? YES. That's exactly how it ought to be – a bonus, not just a date adjustment. I only wish Codex did the same. /cc @thsottiaux
Nova retweeted
I just used a Claude banked reset, and… it doesn't change my main reset date? YES. That's exactly how it ought to be – a bonus, not just a date adjustment. I only wish Codex did the same. /cc @thsottiaux
完全正确，中国大部分软件设计之冗杂繁琐极其糟糕，包括但不限于非要内置浏览器坚决不调用系统浏览器、存储空间占用极大、界面菜单复杂的像是在走迷宫一样。
Nova retweeted
Gemini 4 Argon weekly usage quotas are really low. I'm on the Ultra plan & still ran out of my weekly quota in less than two days in Antigravity, despite using it 10 times less than GPT-6.1 Sol, mostly at extra high to max! I like Argon a lot, but the limits are very problematic.
Nova retweeted
Gemini 4 Argon is a good model but the usage limits are rough
Opus 5.5 users are running it like it's nearly unlimited, OpenAI reportedly just added servers so Sol 6.1 can go unlimited too, and Meta's been pretty generous with Muse Wonder Full
Then there's Google, sitting on more compute than any of them
TPUs were literally built for fast cheap inference
So why are Argon users the ones hitting the wall first
If Google wants people to actually switch they gotta open the tap 🤔
Nova retweeted
Why does Google Gemini 4 Argon have such bad usage?
Claude is basically giving out unlimited usage with Opus 5.5
OpenAI just bought fresh servers for unlimited Sol 6.1
Meta has generous usage with Muse Wonder Full.
Google has more compute than all three of these labs.
TPUs are designed for inference speed.
Antigravity should have the most generous usage.
Nova retweeted
Using Gemini 4 makes you realize how many other models must be distilling Claude/GPT (or using a similar training pipeline). It feels quite different than other models.
