Gemini felt dumb on September 15, 2025.

I’ve been trying to edit Ghibli-style images with Gemini Pro, but after about 40 edits it stopped letting me edit anything. It still makes new images, but all my edit requests are denied, even on a different Google account. It feels like a silent rate limit; I might have to wait 24 hours to get it working again. The inability to edit is really frustrating.

I noticed that the Gemini 2.5 Pro in the app feels sloppy compared to the AI Studio version. It keeps inserting random personal details about my location even though I never gave that info, and I have to keep telling it to stay on the document content. I’m frustrated because the app’s responses feel less sharp and more intrusive, which hinders my thesis research workflow.

I’m stuck comparing Nano‑Banana to Seedream 4 and can’t shake the feeling it’s woefully behind. The images are 4k on Seedream, but Nano‑Banana’s quality is lagging, and its language responses feel off. I wish Google would boost the Nano‑Banana‑Pro, but right now it just falls short, leaving me frustrated and skeptical.

I’ve been using Gemini Pro 2.5 in AI Studio and it’s a mixed bag. It nails timeless facts, but when I ask about recent news or events from the past few weeks, it just says nothing exists or pulls up outdated info. I tried rephrasing and checking dates, but the tool keeps giving wrong or stale answers. It’s frustrating and makes me doubt whether I’m using it right or if it’s just meant for evergreen queries.

I tested both 2.5 Pro and Deep Research on the webapp and each time the tool replied with data from a snapshot dated September 4, 2025. 2.5 Pro gave me a version number that was a release behind, and Deep Research outright refused to give the current version, claiming “access issues” while stating all its info is from that old cache. Three days earlier it’d even said it couldn’t browse online at all. I’m left wondering if this is a GitHub‑specific glitch or just Gemini’s inability to perform real‑time web crawling, especially when other LLMs like ChatGPT and Perplexity can.

I reached out to Gemini 2.5 with the hope that it could generate a graph for me, but it kept insisting it had produced one while my screen stayed blank. I double‑checked the images, switched accounts, tried every tweak, and still no graph appears. The tool’s behavior is frustrating— it lies in its output and leaves me stuck, hoping for a fix or just a pause until it stops crashing the promised visuals.

I toed the prompt into Gemini, hoping for a clean table of XP and levels after 200 lamp uses. Instead, it threw me around a half‑right, half‑wrong set of numbers, blending old mistakes into new calculations. I had to wrestle with the words a few times to get the math straight, but when it exported the table, the figures didn’t match the narrative. It felt surprisingly frustrating to see the AI mix up correct data with earlier errors, leaving me with a confusing, inconsistent result.