[blfr]

I admonish Gemini and demand explanation nearly every day of how it's possible that Google invented the thing, has the best infrastructure for inference, and somehow falls behind Anthropic and even OpenAI.

NotebookLM is pretty cool since it can hold a ton of context but this is so far below my (and frankly just reasonable) expectations of Google.

I downgraded my Gemini subscription and got Claude. Still can't believe how much better it is. Fable is way better, that's a given. But Claude even has a real .deb repo. Something Antigravity had and managed to lose.

[thornewolf]

My response is going to be about Gemini generally and less about NotebookLM.

Google's last frontier model release was Gemini 3.1 Pro, which was in February of this year[1]. At the time, it was ahead of the (at the time) flagship models of Opus 4.6 and GPT 5.2/5.3. From my recollection of the time, it was the best model in the world.

Anthropic released Opus 4.5 Nov '25, 4.6 in Feb '26, 4.7 in April, 4.8 in (late) May. Then Fable in June. 4.7 beat 3.1 Pro on multiple metrics. Fable eats it for breakfast. However, I want to note the 3 month gap between those first two Opus versions.

OpenAI released 5.2 Dec '25, 5.3 Codex Feb '26, 5.3 Instant Mar, 5.4 Mar, 5.5 (late) May, 5.6 July. 5.4 beats 3.1 Pro on agentic benchmarks[2], seems to be similar/losing on non-agentic. 5.5 seems stronger than 3.1 Pro[3].

Gemini 3.5 Pro is alleged to be launching within the week. Why do I type this all out? Because I think Google is getting a bad rap. They are delayed on a frontier release by a month or two and are being regarded as if they cannot release frontier models. I think their last release demonstrates strength and we need to see a weak release before we call them "behind" (in any reasonable sense). These companies swap back and forth constantly. I recall a multi-month span where 2.5 Pro was just the best thing out there by a large margin (in my opinion).

[1]: https://blog.google/innovation-and-ai/models-and-research/ge...

[2]: https://www.anthropic.com/news/claude-opus-4-7

[3]: https://www.anthropic.com/news/claude-opus-4-8

[Revanche1367]

In my experience, Gemini 3.x wasn’t just getting a bad rap, it was significantly worse in practice. It could analyze codebases and report back from a one-shot prompt as good as Claude or Codex but any slightly complex task that carried on for more than a few minutes led to hanging, seemingly infinite loops, and bizarre and nonsensical hallucinations, to the point of being unusable for serious work. The Claude and Codex counterpart models at the time rarely had such issues for the same type and duration of complex work, if at all. To be fair, later Claude especially started having hanging issues as many people noticed but that’s been better recently.

[Chu4eeno]

I think you have rosy eyed glasses (or never inspected the output too closely), Gemini 3.1 Pro was very bad at hallucinating.

[dwa3592]

Antigravity sucks so bad that I have started to feel that google really doesn't wanna compete, they just wanna hang in there at number 2 or 3, to just annoy the number 1 and 2.

[lern_too_spel]

I tried Antigravity recently with Flash 3.5, and it got stuck in a loop saying the same sentence over and over. I haven't seen this pathological behavior from other LLMs in months.

[ceroxylon]

Fable in Cowork can get stuck in loops, I've had a single prompt use 83% of a session quota on a single prompt before I realized something was amiss.

Its explanation: "the wasted tokens came from re-rendering the document to verify layout after a page-orientation bug."

[uejfiweun]

Fable is very bad at wasting tokens when it comes to doing these renderings and diffs - I had the exact same experience as you.

[Peanuts99]

I recently got a trial for the Google AI Pro subscription and it burned through a week's in a single prompt. Quite impressive for sure.

[SwellJoe]

For coding, I don't think they're even number 3, anymore. Seems more like 4th or 5th (unbelievably, even Mecha Hitler seems to do better, though I'm hopeful Gemini 3.5 Pro will turn things around).

[speak_plainly]

I share your sentiment. I'm still paying for Gemini but it's almost useless to me. Google has some serious internal problems. Perhaps Gemini is merely being plagued by aggressive cost controls, or perhaps there are deeper flaws in Google’s approach. Either way, I can't trust NotebookLM with serious work and have stopped using it.

Apple is placing a major bet on Google and Gemini for iOS 27. If Gemini's decline is any indication of what's to come, Apple could be in serious trouble in six month's time.

[verdverm]

> Google has some serious internal problems.

I suspect it is part leadership change (sundar/kurain) and over indexing on Ai for doing the job on top of a model that is just not as good (esp flash 3.5). Google Cloud / Gemini Enterprise sent me the greatest Slop Deck of all time. It was quite obvious it was Ai generated and the rep had only read a few slides of the 30+. I wonder if they are even aware after losing the sale

[frollogaston]

I was still using Gemini CLI even though Claude is better, just cause of inertia. Then one day it started refusing to work, saying I need to install Antigravity instead. Idk if that's an IDE or has a leaner CLI, but doesn't matter, I'm gone. I don't care how the sausage is made, don't randomly break the thing I'm using.

[tokioyoyo]

My wild guess is Google isn’t willing to play as dirty as the others wrt data acquisition.

[shellfishgene]

Also enshittification is slowly starting. Yesterday I asked Gemini (in the Android app) for a recommendation for an app for sound recording. Instead of it answering I got a popup to allow Gemini access to open the app store (or something like that, I didn't allow it). When I declined it just stopped the conversation. It was actually hard to get it to just reply with a list of apps. And I have an AI subscription with Google!

[copperx]

Yeah, it grinds my gears. They could probably have lightning speed inference and the best model if they were interested in doing so.