I was glad to have the AI assistant back, but now new issues have come up. First, the AI takes way too long to process requests, and it feels like it's overthinking simple tasks. Second, I keep getting the error: 'AI hiccup — retrying…' multiple times in a row. This is highly frustrating and severely slows down my development workflow. I assume the servers are either overloaded or there is an internal bug with the AI. The assistant itself mentions it's unable to generate responses. Apologies for the tone, I'm just really frustrated at this point lol. Is it just me or is anyone else experiencing this? Let me know. I think it’s not a major issue, but it’s still super annoying. Ps: (Turns out it’s a model from the Muse Spark family just "Muse Spark," not the 1.1 version or anything like that I don't know if that's the issue, but it still feels a bit sluggish to me.)

[–] [S] 2 points 1 month ago

It is great that you are willing to support the project financially to bring the old model back. But if they introduce mandatory sub fees, the whole meaning of this website will be lost. No matter how much we want the smart model back, Perchance was always meant to be a free digital canvas. It is like having a brush and paints that you borrowed from a friend to create art. Anyone should be able to just come here, build something for fun, and share it with the world without a price tag. Turning this place into a commercial service with paywalls would completely ruin its soul and its freedom.

  • source
  • parent
  • context
  • [–] 0 points 1 month ago (1 child)

    You made some fair points, and I agree that "take what you get" is the reality right now. But honestly, this bad. The whole point of having an AI "assistant in a chat" is to just talk, share an idea, and build things together. I want to communicate, not spend half my time engineering massive prompts, micromanaging its steps, and pointing fingers at every single mistake like I'm babysitting a ten year old text generator. Yes, we can adjust, but we shouldn't pretend that turning a smart assistant into a literalist machine that requires manual guiding is a fun experience. But please don't get me wrong, I'm not just here to criticize. I understand that the developer had to do this because of server costs, and it is a tough situation. I really appreciate everything the dev does for this site, so I guess we all just need to be patient, stop whining, and wait for better days.

  • source
  • [–] [S] 2 points 1 month ago

    Thanks for the link. I see now that it was an emergency downgrade due to server costs. Glad to know the developer is looking for a way to bring a smarter model and the thinking text back. We'll just have to wait.

  • source
  • parent
  • context
  •  

    This post is going to be a long one, filled with my personal experience and comparisons because I really need to get this off my chest. And yes, I’m a newbie here. I discovered the site 2 months ago, though I only registered this account recently.

    As I think other Perchance users have noticed, the AI assistant has been acting incredibly dumb lately. It constantly glitches, writes in a dry, unpolished, and overly brief way, delivering the absolute minimum. It completely lacks any 'imagination.' Unlike the previous model, it can’t brainstorm ideas based on your project anymore; it just repeats what was already written in its readme file. It’s completely incapable of coming up with anything original, or at least it struggles heavily to do so.

    I really loved how the old model used to write, highlighting and noticing every little detail. Heck, I even experimented with it, and the model could synthesize sound or literally copy a song, which is already amazing by itself, or edit audio. Sure, it can't actually 'hear,' but still! All of this is from my own experience. But if you compare it to the current model... well, I’ve already described it: it's dry. It is fast, that’s a fact, but I would much rather wait longer for the sake of quality.

    So, what I’m getting at is that I hope this is just a temporary workaround by the developer. The website is great, and the AI agent used to be a massive help. It brought ideas to life. It was awesome to build a game framework and then just polish the minor details. It wasn't perfect, but the quality was high.

    What makes it even more frustrating is the 'thinking' feature. When I first saw it, I wondered why it was even there. But then I realized it’s a brilliant feature. Now, though, I don’t see a single trace of actual thought from this dummy.

    Anyway, I’ve rambled on for too long. I hope this doesn’t last. I can see that people are furious because, let’s face it, 'how could they take away our gold and give us copper instead?' But as harsh as it sounds, we also need to stop whining. No matter how disappointed I am, the developer might be having a hard time or dealing with temporary issues, so we probably just need to wait. Everything is so messy right now.

    I might be inaccurate on some points, so please take this with a grain of salt. It's just my view on the current situation.

    [–] 3 points 1 month ago

    I agree I thought I was the only one, but it seems to have affected practically everyone it’s all pretty complicated. And that’s not even touching on the design aspect: the old model excelled at sculpting and icon generation it was reliable. Here’s my experience: the AI ​​agent used to understand exactly how a character should hold an object and what the texture should look like. Take a pickaxe, for instance the old model knew how the character should hold it, whereas now, if you give the new model the same instruction, it might end up placing the pickaxe in the hand upside down. Can’t the system double-check that?

  • source
  • parent
  • context
  • [–] 4 points 1 month ago (3 children)

    As difficult as it is to accept, we need to adapt to the current situation. A high-quality AI agent won't be available in the foreseeable future, so our best option is to work with the tools currently at hand

  • source
  • [–] 1 point 1 month ago (1 child)

    I still don't get it, I can't see his thoughts. If that's what you're talking about, just forget it it's a standard glitch. I had something similar happen back when the AI ​​agent was working fine. I hope they bring back the proper AI if this is just a temporary issue, the developers are understandable I just hope it's not a case of them slacking off.

  • source
  • view more: next ›