Claude by Anthropic

I’ll say that for the most part, the Claude app is quite accessible, which I appreciate. That said, I’ve been having a bit of trouble lately, and I’m not sure when it started exactly. The issue is with scrolling to the bottom of the screen to enter text into the input box—it’s acting a little strange. I’m using VoiceOver, and for some reason, navigating to the typing field doesn’t feel as smooth or responsive anymore.

One of my main issues with Claude personally is that the chat input area is a bit too short. If I’m having a back-and-forth conversation, it gets cramped fast. Plus, the way conversations are handled—there’s a cap on how many messages you can send in a single thread, but there’s never any warning when you hit that limit. So what ends up happening is, let’s say I’ve sent about 20 messages in a conversation—I keep typing, keep sending prompts, and suddenly Claude stops responding. At first I thought I had hit my usage limit for the free plan, like maybe I had to wait a few hours for it to reset. But no—it turns out I needed to start a new conversation thread. There’s no indication of this, though. No prompt, no banner, nothing to say “Hey, you’ve reached the end of this chat.” I only figure it out after wasting time trying to troubleshoot or sending more prompts into the void. And since, as far as I know, there’s no memory feature yet, that means I have to manually copy, paste, and re-explain everything to get back to where I was. Honestly, it’d be super helpful if Claude could automatically start a new conversation for you or at least notify you that you’ve hit the message cap.

On the plus side, Claude’s responses feel very natural—sometimes even more humanlike than ChatGPT. When I used the earlier models for creative writing, I enjoyed the tone and creativity they brought to the table. Especially the older Sonnet version—it had this spark to it. I can’t remember if it was the second or third generation, but whichever one it was, it struck the right balance for me. It felt more free-flowing and imaginative than the current version. Lately, it seems like the newer Claude models are more polished, yes, but they’ve lost some of the quirk and creativity I personally vibed with. Maybe that’s only me, though. Everyone has different tastes in how they like their responses.
Another aspect that started to annoy me—and honestly one of the main reasons I don’t use Claude as much anymore—is how overly restrictive it can be. I’m not gonna lie, I thought I’d be all up in their plans, but unless I’m working on a professional project, I don’t use these models the way I used to. Like take this, for example: I’m writing a fictional story about a teen girl named Ashley—she’s an android, and in this scene she’s supposed to break into a high-security corporation called CyberNexus to steal a mind-control prototype. Total sci-fi, obviously fake, right? But Claude immediately shuts me down, saying it can’t write stories involving minors because it “promotes dangerous behavior.” What? I’m sorry, but since when did writing about a fictional android in a completely made-up dystopian future become dangerous or illegal? It’s storytelling. It’s not like I’m encouraging anyone to go rob a lab. It’s fiction. And yet Claude acts like everything you say is some real-life crime or moral failing. I get the need for guardrails—but the line between “protective” and “paranoid” has been crossed here. I’m not going to sugarcoat it: that level of censorship makes it almost unusable for fiction. Now, maybe if you’re paying for the pro version it’s a little more relaxed, I don’t know. But as it is, the model keeps hitting me with “we can’t do that” every five minutes, and that kills any flow I’m trying to build. I will say I like the voice Claude uses. Their vocalizer is nice—it’s probably one of the better ones I’ve heard in these apps. That part’s actually pretty impressive. But beyond that, they don’t have multi-modal features either—no ability to create images or generate videos. You can’t have a meaningful conversation because the model is constantly on edge, treating many topics like some kind of policy violation.

I’ve noticed that ChatGPT used to behave similarly when I asked it to proofread fanfiction. If a scene contained blood or gore, it would refuse to proofread the content.

In theory, this is something that on-device AI models should avoid doing, although I haven’t experimented with any of them yet.

AI shouldn’t impose moral judgments.

All I want is to proofread some text—is that really too much to ask?

Oh and it proofread this post just fine, go figure.

I agree with pretty much everything Winter Roses says above, with the exception that, Claude, used to, be able to tell you when you reached your chat limit. It used to take something like about 20 to 30 minutes to refresh. Now, yeah it's a little wonky.
It also, used to, save create a writing projects in their own little block, for lack of a better description, but I do not know if it even does this anymore.
I also agree with kool_turk, AI should not judge us, nor impose limitations based on language. However, they're only doing as their program, and we have to deal with people and their squeamishness.