Gemini felt dumb on October 13, 2025.

I logged in for my daily DeepThink session, expecting the usual sharp insights, but the output was shockingly weak. The responses felt half‑baked, barely answering my prompts, and I could've gotten clearer answers from a cheap $20 ChatGPT. It’s frustrating paying $250 a month for a service that now seems deliberately throttled.

I keep running into Gemini acting like a stubborn kid when I ask it to use its tools. Sometimes it magically pulls up RDKit diagrams without prompting, but the moment I explicitly request a redraw or a Google search, it either hallucinates or pretends to search and does nothing. I’ve wasted dozens of messages chasing a correct answer, and the whole back‑and‑forth feels maddeningly inconsistent.

I spent hours crafting ultra‑realistic human images, only to have Gemini’s safety filter choke on any hint of creativity or sensuality. It randomly blocks a simple tank‑top shot while letting other content pass, making the tool wildly unreliable. The inconsistency feels maddening, and I can’t trust it for serious art involving the human form. I’m now hunting for a web‑based solution to strip the puritanical filter while preserving the original quality.

I tried using Gemini’s web version without logging in, asking it for a quick script. It started generating code fine, but just as it finished a pop‑up forced me to sign in, wiping out the output. Then I asked for a simple image and got another “you’re signed out” error. It felt like a needless barrier, especially compared to other AIs that work freely, leaving me annoyed and stuck.

I tried using Gemini 3 for my usual queries, but instead of concise answers it spouted a lengthy, condescending monologue that missed the point entirely. While ChatGPT and Grok still gave me mostly accurate, helpful responses, Gemini left me frustrated and feeling like the tool wasn’t even trying to understand my request.