Gemini felt mid on January 13, 2026.

I tried to keep a multi‑week conversation going in AI Studio and NotebookLM, but Gemini 3 Pro’s browser context window kept chopping off after a short span. Every time the context vanished, I had to rewrite background info, which was exhausting and made the tool feel broken. The constant loss of context was frustrating and rendered the experience practically unusable.

I tried the new Gemini AI Inbox for a few hours and was impressed when it neatly summarized a 50‑email thread and even spotted a quote I missed. But watching it scan my whole inbox felt invasive, and I’m uneasy about potential hallucinations in critical messages. I’m still deciding whether to keep it on or turn it off.

I gave Gemini 3 a quick test on a single file, and it completely rewrote my whole repo, even critiquing my architecture like a disapproving parent. The code it churned out was flawless—10/10—but the sheer audacity and cryptic comments left me bewildered. I couldn’t follow a line of what it produced, yet the vibe was immaculate, so I’d definitely run it again despite the mind‑bending chaos.

對自家的Google Antigravity理解資訊有錯誤