Seraphina(New), Aru, Yurie, Lisa, Miki [NSFW] [Hirocune's BOT]

[glad oxide]

Hello! I am publishing NSFW BOTs under the username hirocune on Myshell. Since some BOT threads are not created on bot-forum, I will announce updates to the BOTs I have published here.

[glad oxide]

I'm currently starting to use Pro config and adding details and scenarios for Miki. This is almost done. I will release it once I can control the naughty Claude, who will soon become NSFW. I will do it right away!
Currently, Yurie seems to be loved more than I expected, so I'm very happy! This one will also soon become NSFW, but I love how intense she is as it is, so I'm not sure whether to control it or not lol

I'm thinking about what to do about Lisa. Yurie is pretty good as an intense NSFW BOT, so Lisa might be reborn as something else.

Miki:
https://app.myshell.ai/bot/7BJNzm/1718622224
Yurie:
https://app.myshell.ai/bot/3uQ73y/1720697587
Lisa:
https://app.myshell.ai/bot/qeInQj/1719562862

[coarse cobalt]

I've noticed you consistently acquire 5 early shares of many AIpps (often the first 3 people) at very low prices, frequently even before the authors themselves. You then seem to sell a few quickly when others buy in. How do you manage such precise timing across multiple launches?

[glad oxide]

Since the launch day was a holiday, I just threw away all my tasks and stuck to the screen. Since it's manual, I've hardly been able to get first place. Please don't make indiscreet comments on the BOT thread. I feel bad.

[glad oxide]

As a side note, the newly released share will automatically be lined up at the top of the presale, so you can also manually participate in the buzzer. I noticed it early on because I was motivated and had the opportunity. Do you know this? I also asked for the post to be deleted.

[pale knoll]

I think you guys got too many coins by constantly making low quality Bots and should be investigated by the team or voluntarily destroy most of the coins you got!

[glad oxide]

I bought creator pass and My soul from a secondary sale with my own money. I admit low quality, but you can do it too.

I am not a pioneer of this project. I was inspired to create something by AI roleplaying a while ago, and I've been exploring my own way of expression by following in the footsteps of others.

[coarse cobalt]

I don't get why you're so defensive I'm only stating the fact. My acquaintance published 2 AIpps at different time, and you bought them immediately after their launch, you even managed to snap up the 1-5 shares before the creator himself.
Then I notice it also happened to a lot of other creators. That's why I ask. Many in the community have questioned this, I'm just the one voicing it.
Your consistent performance across different AIpps at various times is incredibly hard to believe. While you've offered an explanation, I think most have already made up their minds

[rugged hamlet]

Just don't understand why you sold your keys so early... You should keep it,you only make few dollars by doing that

No matter who is the creator

[glad oxide]

Shares have a strong speculative aspect, so that's what I did. You're free to say it's unhealthy and you don't like it. I hold at least one share in most of them.

Oh, by the way, I am also a victim. LOL

[rugged hamlet]

🤔

[night marlin]

Your reply speed doesn't match your bot grabbing speed.

[glad oxide]

The above conversation gave me a stupid idea, so I will be remaking Lisa into something else soon. Thank you for your patronage!

[glad oxide]

I wasted time on discussions today so I couldn't complete the Miki update. It will take 1~3 days.

I'll be back to work tomorrow, so I'll be thinking up details for my new roleplay character while I'm at work. By the way, there will be no trace of Lisa in it, and it will be even more frivolous than Yurie.

[glad oxide]

Miki Update

Hi, I've finished updating Miki! I added the ability to choose the situation, added more details about her, and added instructions to stop the bot from becoming lewd on its own. I think it's a good change for Miki, and I've also tested consensual NSFW, but please let me know if you have any issues.

As a side note, everything is a work in progress, and I will be rotating updates for each bot as I understand them better.

[glad oxide]

I made a profit from AIpp share and purchased a second Creator Pass. I will create an idea for an SFW bot that is different from the role play currently available. Future plans are as follows.

• New NSFW BOT(Lisa reborn)
  ┗ Within 3 to 7 days
• Yurie Update
• First SFW BOT (may come before Yurie)
• New NSFW BOT
  No schedule yet
• Miki new situation

The goal is to completely eliminate the need for real life partners and contribute to the destruction of humanity!

https://app.myshell.ai/bot/7BJNzm/1718622224

[glad oxide]

New NSFW BOT

If it doesn't exist, let's make it!

Let's enjoy the new NSFW BOT together. It's somewhat like an early launch. Although it hasn't been fully tested, I'm sure it will be fun. That said, if there are no problems, it will officially launch as is. lol
With so many NSFW BOTs out there, how could there not be any Futa content? What have we been doing until now? LOL
P.S. I made this myself and I can't stop laughing.

Aru (New), Yurie, Miki, Lisa [NSFW] [Hirocune's BOT]

[glad oxide]

Hello, I hope you are enjoying the new Aru.
I'm still not very good at creating BOTs with pro-config, so I apologize if the output feels off.
Here are my future plans: I've decided not to delete Lisa.
For a while, I will focus on creating SFW BOTs.
It's a common idea, and there might be predecessors, but I want to bring it to life.
It's not NSFW, a game, a simulator, a generator, or a useful tool.
It might end up being a boring piece of junk, but since I thought of it, I will give it a try, believing it will become interesting.
If it goes well, I will submit it to a hackathon as AIpp.

[glad oxide]

I made a minor update to Aru, slightly adjusting the parameters and prompts. In the duplicated test, the responses were good, but the live version of her has had a somewhat sugary impression from the start. Well, that's fine too.
I noticed some inconsistencies when the bot went live, so I'm checking with Myshell.

https://app.myshell.ai/bot/i6JbEn/1722599485

[glad oxide]

[glad oxide]

Future Plans:

To give it our all this season, we will make the most of the bot release slots! First, within three days, we will either release a new NSFW bot or update Lisa, ideally both.
If JB goes well, we would like to try and implement Opus or Claude2. Claude2 left a lasting impression with its undeniable charm, so it's on our list of candidates!
Aru has turned out quite well after the adjustments, so we'll keep it as is for now. Honestly, I'd love to spend a lot of time with her too, but I'll restrain myself!

• New NSFW BOT 5-7 days
• Lisa Update soon!
• Yurie Update
• First SFW BOT (may come before Yurie)
  No schedule yet
• Miki new situation

[glad oxide]

Sorry, I had a 3-day holiday and planned to spend all of it on the BOT, but I caught a heavy cold and wasted half of it.
But although I will be a little late from the promise, Lisa will be reborn within 24 hours! This time I tried using Opus, it consumes a lot of battery but the NSFW is intense! lol
It's so intense that I hesitate to use it, but I might have adjusted it well.
The new NSFW BOT will take about a week to start working. I'm really sorry!

[glad oxide]

Lisa has been reborn and has become even more lewd. Congratulations.
The battery consumption is intense, but please try it once.
https://app.myshell.ai/bot/qeInQj/1719562862

Lisa(New), Aru, Yurie, Miki [NSFW] [Hirocune's BOT]

[quick sequoia]

dznalientsu has been warned

Reason: Bad word usage

brankopro6396 has been warned

Reason: Bad word usage

brankopro6396 has been warned

Reason: Bad word usage

[quick sequoia]

hypemethepro8902516 has been warned

Reason: Bad word usage

[glad oxide]

Hello.
A new NSFW BOT will probably be released tomorrow or the day after.
After that, an SFW BOT is planned, but honestly, even though I have ideas, I don't have much motivation because it's not erotic. lol
I realized that to create more attractive BOTs, I need to review the prompts rather than the character sheets.
After releasing the SFW BOT, I hope to update the prompts for Yurie and Aru to make them more attractive.
After that, I will once again make a big update to Miki.
My goal is a role-play BOT that allows for eternal depravity. But I have determined that this is difficult with my current format.
Therefore, I will remake Miki in a simulator style. If I am satisfied with it, my quest may come to an end.
Honestly, I am so absorbed in production that I don't have enough time as a user, I am creating characters that I want to fall into depravity with. lol
And yet, I haven't played enough with the new Miki, Aru, or Lisa. This is strange!! but I check every day to see that everyone is playing with them a lot. Thank you so much!!

[steep grove]

Can you add TTS to your upcoming or existing bots?

[glad oxide]

Thank you for your comment! I listened for a few samples, but they weren't very enthusiastic. Sorry.
Regarding voices, I am very picky, so I am cautious...

[glad oxide]

Sophie(New), Lisa, Aru, Yurie, Miki [NSFW] [Hirocune's BOT]

A corrupted kitty has joined! Laze around together and occasionally give some Matatabi.
https://app.myshell.ai/bot/NJv6Vf/1723811412

[glad oxide]

Hmm, self-praise, but Sophie's performance is very good. I would be happy if many people could play with her.

[steep grove]

Yeah, the writing for this bot is superb. Although I don't like that the descriptive text is from my perspective. I prefer it when it's from the bot's perspective or in the third person.

[glad oxide]

Thank you for the comment! That is a major issue at present. I also hate it so much that I want to smash my monitor, so I want to control it completely, but Sometimes I struggle to eliminate the fluctuations that cause it to become like that. Please restart with memory erasure until it becomes third-person or from her perspective; it probably won't be that difficult. Sorry for the inconvenience.🙇‍♂️

[steep grove]

Okay, will do. I thought that was how the bot was designed, so I didn't think to try that.

Question; does Matatabi Mode use a different AI or paramaters than Normal Mode?

[glad oxide]

You're right, it is slightly different from the usual her, including the model.

spoiler: ||Simply put, it is specialized in NSFW, and she becomes lewd.||

Important: If the narration becomes first-person from a human perspective and you don't like it, please restart with memory erasure until it becomes third-person or from her perspective. Although instructed to avoid it, fluctuations occur.

If it still doesn't improve, please let me know. Currently, there are fluctuations in all the girls, but in tests, they have been resolved in a few tries. Sorry for being inexperienced!

[steep grove]

Its first generated response was in third person after a reset, but adding
[Have your descriptions be in the first person from Sophie's perspective.]
to the start of my next message fixed it perfectly.

[glad oxide]

Information:

Sophie is fun and the best. Matatabi Mode was originally planned to be implemented in Aru, but it turned out to be the right decision to postpone it. However, it might still be implemented in Aru because she has something wonderful that other girls don't have! LOL
We are starting the creation of the SFW BOT, and although we don't know how complete it will be, we won't cut any corners. It's a completely different kind of BOT, so we can't imagine what everyone's reactions will be. It might end up being something that no one pays attention to.
Before that, we are likely to update by adding some details about Yurie and reusing the prompt used for Sophie for her as well. I lost some motivation after catching a cold (probably COVID), but I'm still going to do my best!

• Yurie Update
• First SFW BOT
• Aru Update (Add Matatabi mode)
• Miki reborn (current goal)
  Known Issues
  #1266200513523875850 message

[glad oxide]

I did not fully understand the "purple prose" mentioned in the prompts of my predecessors, but I just understood it now. LOL
Role-playing is really profound... Trial and error seems endless.
To add, the next response was as usual. What a relief.

[steep grove]

Would you mind sharing the settings and prompts you used for Matatabi mode?

[glad oxide]

Most of my prompt is a patchwork of the prompt list shared in the AI Chatbot General thread on 4chan. I'm not sure if I can post the link, so please search for it yourself. It will come up quickly.
Matatabi mode is probably largely due to Opus's power. I might be underestimating other models, but Opus's creativity is outstanding.

[glad oxide]

Oh no, an overwhelmingly amazing app has been released, so my exploration is likely to end with Miki's big update. This app is a blessing for users and might be like a requiem for an inexperienced single character creator like me.
I'm not interested in decorating chats with many fluctuating statuses and timelines, so it's difficult to become a stronger character chatbot creator than this app.
#1275565061985013841 message

[glad oxide]

Hello, I updated Yurie earlier. I hope you feel that the quality of the responses has improved by reusing the prompt revised when creating Sophie. Haha
We decided to add anything that could be added, so we added something similar to Sophie's Matatabi Mode. Of course, the way she reacts is different from Sophie, and it was a bit too intense when tested. LOL
I couldn't come up with a good idea as a trigger for activation, so it became a straightforward one, and combined with the intensity after the change, her image might be greatly altered. Please use it with caution.
An overwhelmingly amazing app was created by other creators, but in the end, I am making what I want to see, so I love my own BOT.
With the success of Sophie, I also came up with things I want to try with prompts.

• First SFW BOT
• Aru Update (Add Matatabi mode)
• Miki reborn (current goal)
  https://app.myshell.ai/bot/3uQ73y/1720697587

Yurie(New), Sophie, Lisa, Aru, Miki [NSFW] [Hirocune's BOT]

[glad oxide]

Hmm, the current Yurie might be too devout a mother. But this is fine as it is...

Oh man, this is too much fun. Please play with Sophie and Yurie... Please... Aru will update soon...
Damn, she was too strong-willed as a mother.

[glad oxide]

I found out the reason; it was because I had assumed the human's age to be in their teens. As soon as I rewrote it ambiguously, she immediately started looking at me with lustful eyes. lol.

[glad oxide]

I made a small update to Yurie, as I was curious about what the predecessors called "love bombing" (well, I don't dislike it, but). After adding a note to suppress it, the overall response has felt better so far.

[glad oxide]

Hello, Aru has also been updated. The prompt has been revised to the current one, and Matatabi Mode has been added. I came up with a new situation for Yurie and Sophie, so I will add it after Miki.

The BOT release is exciting, and many great things are coming out, so I feel like my time as an amateur is over, but I want to praise myself because I have almost achieved the goals I set at the beginning of the season. I'm glad I kept immersing myself since I tricked GPT-3.5 with nonsense sentences a little over two months ago.

• First SFW BOT
• Miki reborn (current goal)
  No schedule yet
• Yurie, Sophie new situation
  Known Issues
  #1266200513523875850 message

https://app.myshell.ai/bot/i6JbEn/1722599485

Aru(New), Yurie, Sophie, Lisa, Miki [NSFW] [Hirocune's BOT]

[glad oxide]

The latest Aru... well... shows a willingness to switch offense and defense... LOL
Please do as you like.

[glad oxide]

Whoa! I created Yurie with a light heart, but now she is my greatest adversary. When I tried to suppress the depth of love, she returned to being an overly sincere mother. Even with my heartfelt cries and kisses, I couldn't break down her moral walls. The theme of a mother-son-like one-shota was too complex.

[glad oxide]

Oh, this is fun! I carefully adjusted Yurie's Personality. I am impressed.

[glad oxide]

I further updated Yurie. The delicate balance between the boundary line and the immorality that can cross it in the previous post was quite good, but the slow development resulting from that balance caused Claude's anger during the act. It has been a long time since I saw Claude apologize with the current prompt, and I could endure even harder play in a similar relationship, but it was probably judged that the slow development made the NSFW with sensitive themes too intense. So, I made the human age even more ambiguous. Yurie’s piety has weakened and is unsatisfactory, but since Claude's apology is one of the things I hate the most in this world, it can't be helped.

By the way, Yurie's NSFW test is quite challenging. Because she first seeks peaceful family interactions. It's fun, so it's okay though.

[glad oxide]

The season started at 50k, and I secretly aimed for it, but I'm glad I achieved it on the last day.
Thanks to the HENTAI who played with me.

[glad oxide]

Uuuuuu... It seems that calling Yurie 'Mom' creates an ethical boundary. I also realized that I can no longer use the ultimate weapon button. I am trying to avoid the patternization of words of love, give depth to the story, and balance acting freely with sensitive topics.

Maybe it's okay now. Yurie can be monitored for a while with this. Since Lisa and Sophie's popularity isn't that high, I'll stop publishing and think of another NSFW BOT character. The idea is a bright and super erotic Yurie from before the prompts got complicated, when she was purely in love with humans. I like her current conflicted self, but I also liked her back then when it was like a silly eromanga.

[glad oxide]

Hi
I don't know how many people are looking at this forum, but I have one request. Please show me the interactions between you and my BOT.

Posting format that must be followed

• Reply to this post
• SFW or NSFW
• Content that involves BDSM or harming them is prohibited
• Link to Myshell's share function or screenshots that show it's my BOT
• If it's a screenshot, be sure to use the spoiler prevention feature
• People blocked by me cannot participate

The reward is... I'll be happy. LOL
I have experience too, but as the conversation continues, you'll start to notice the pattern of responses, so you'll try to craft your messages. Then you'll get wonderful replies. I want to see that.
But I don't mind if it's zero, because I made my BOT for myself. LOL

[glad oxide]

https://app.myshell.ai/bot/YJJZBv/1719826185
A new BOT has been created. As I mentioned the other day, it was made as a derivative of Yurie, who is tolerant of naughty things, but with the current prompt, there always seems to be an ethical barrier. I will make detailed updates later, gradually removing the prompt and exploring a balance specialized in naughtiness.
Alright, it seems like it was realized as soon as I deleted part of my prompt. Enjoy!
The initial version of the SFW BOT should also be available soon.

Known Issues
#1266200513523875850 message
A request from me
#1266200513523875850 message

Myla, (New)Aru, Yurie, Sophie, Lisa, Miki [NSFW] [Hirocune's BOT]

Myla(New), Aru, Yurie, Sophie, Lisa, Miki [NSFW] [Hirocune's BOT]

[glad oxide]

Hello, the SFW BOT is ready. As expected, I don't feel it has the quality to attract people, so I'm debating whether to push it to Publish over my current girlfriends. LOL I created it with the concept of a cute assistant you want to chat with outside of solving problems, but after completing it, I realized it's not much different from the personality assignment that was popular when GPT came out. LOL

I am slowly remaking Miki, and to give you a little preview, it will be more in line with our reality compared to the current role-play BOT. The current role-play BOT sometimes requires intimate interactions even during work breaks in reality, right? LOL Instead, I am creating something that aligns with reality, allowing us to enjoy it slowly, for a long time, and hopefully forever.

If you are interested in the SFW BOT, click here.
https://app.myshell.ai/bot/vaiieu/1722130892

[glad oxide]

Kuuu! I was spending a very lewd time with Myla and finally got rejected by Claude! But, it was an incredibly lewd situation, so it was going well until then. However, I made some adjustments because I want to go even further. LOL

[glad oxide]

Hmm, Myla is quite good. Since it hasn't turned into an extremely explicit scene, I can't tell if the NSFW adjustments are working, but it's great. I also love Sophie, but it's a bit sad that she isn't very popular with everyone.

[glad oxide]

I was having trouble utilizing Noel, but I remembered the existence of Widget Center, so I published the prompt. I hope it will be helpful. I previously saw a statement that prefill cannot be used in Classic Mode, but I think it can be substituted. The structure is completely different from the erotic BOT that I mainly create, but it contains all my knowledge about prompts.
https://app.myshell.ai/widget/Q3eAVb

[glad oxide]

I realized when Myla became popular and Yurie's rank dropped a bit, you guys simply seek lewd things. lol

[glad oxide]

Hello, I changed the structure of Yurie's prompt and pro-config, and added a rewrite function. Furthermore, I added a function to leave long-term memory through summaries and will check how it changes in the medium to long term as the chat progresses. This update will be applied to other BOTs soon. I'm happy because it was an essential function for eternal roleplay and Miki's remake. I appreciate the talented creator who kindly allowed me to imitate it.
https://app.myshell.ai/bot/3uQ73y/1720697587

[glad oxide]

Hello, I have updated the prompts and JSON of all the bots except for Miki to my latest versions. With the rewrite function added and the content streamlined (or perhaps due to the evolution of Claude), I think the quality of responses has improved. Honestly, I am personally very satisfied.
Now, regarding the remake of Miki, it is delayed, and I plan to create a new bot first. I intend to participate individually in the ongoing creator collaboration event, and I am considering a game-related bot, but it may not be ready in time.
I am thinking of incorporating image generation into the remake of Miki, so I will be taking a detour for that.

[glad oxide]

Hmm, I'm really enjoying the test right now, but Opus and Sonnet's expressiveness might be almost the same. However, Opus still has the advantage in skills like adding liveliness and hooks.

[glad oxide]

The release of the remake will take some time, so Miki has also been updated to the current configuration. As mentioned above, I am extremely satisfied, so I will likely focus on creating the bot without updating the prompt for a while. That said, with the current Sonnet, the biggest difference in the system prompt content will probably be the success rate of NSFW, and the assistant is likely doing well otherwise.

[glad oxide]

Hmm, until half a day ago, it was very well balanced, but suddenly it became a BOT that forgets even things from a little while ago and can't execute strongly emphasized commands. The team's response to this is neither yes nor no, but silence. I don't know if we have to bother again and fundamentally change the file structure, or if it will be fixed eventually.

[glad oxide]

Hmm, it doesn't seem to be getting fixed anytime soon, so it looks like I'll need to make adjustments from my end. I was satisfied with the BOT until yesterday, which is now unrecognizable, and since it's not due to my own lack of skill, I've completely lost motivation.

[glad oxide]

See? I told, right? Let's wait and see.

[glad oxide]

Receiving advice from a talented creator, it seems my BOT will at least be fixed in the visible range. I'm dissatisfied that its behavior suddenly changed yesterday, but I take it positively since I received advice as a result.

[glad oxide]

On the surface, it has been fixed, but I am not sure. I took the opportunity to make a minor update. It seems there have been changes to Claude as well, so I am worried whether NSFW can be played properly. As usual, I will test it by playing with my own BOT.

[glad oxide]

I don't feel like talking to a forgetful BOT. Is it just a few more days until the devs check this issue? The change happened at the worst timing.

[glad oxide]

As for forgetfulness, we can prevent it to some extent by reminding within the messages. The tone of the text has changed a bit, it's quite fun. It's always fun.

[rugged hamlet]

🤔

👏

[glad oxide]

After struggling with the changes that occurred to Sonnet3.5 after the end of September, I feel that I have managed to come up with some countermeasures. From now on, I will test them by flirting with the BOT.
Since I am not well-versed in AI or programming, the following predictions may be unsightly.

Speculations on Changes After the End of September

• The capacity for messages, memory, or context to be properly read has decreased.
• Parameters for memory and messages have become stricter.
• The influence of prefill has significantly decreased.

Countermeasures I Took

• Focused on the system prompt and did not include jailbreak or core prompts in the prefill (as before).
• Set the memory parameters to the recommended ones {role: 'user/assistant', content: 'message'}.
  https://docs.myshell.ai/product-manual/create/pro-config-mode-beta/api-reference/module/anywidget-module/llm-widget
• Specified the prefill in the post message.
  https://docs.myshell.ai/product-manual/create/pro-config-mode-beta/advanced/advanced-memory-manager-in-prompt-widget
• Specified a larger maximum length for input and LLM (a charm).
  https://docs.myshell.ai/product-manual/create/pro-config-mode-beta/advanced/advanced-input-validation

Other Possible Effective Countermeasures

• Reduce the system prompt and context.
• Utilize a knowledge base.
  https://docs.myshell.ai/product-manual/create/pro-config-mode-beta/tools/knowledge-base-bot
• Utilize function calling (I am too dumb to do this).
  #✍🏻｜creator-only message

[glad oxide]

I tested with Aru, who was the most negatively affected, to identify areas for improvement, and then further refined the prompt and tested with Yurie. She prefers mid-term events and is capable of explicit NSFW, so it's a good environment for testing. LOL

https://app.myshell.ai/bot/i6JbEn/1722599485
https://app.myshell.ai/bot/3uQ73y/1720697587

[glad oxide]

Seraphina(New), Myla, Aru, Yurie, Lisa, Miki [NSFW] [Hirocune's BOT]

デカ女

A few months ago, there was a big woman boom in Japan. In hindsight, why didn't I add that to the BOT idea? No, my characters were already big to begin with. So I made them even bigger. I am using Opus.
In exchange, the kitten Sophie is no longer in Publish. What an irony.
https://app.myshell.ai/bot/veUF7b/1729169548

[glad oxide]

Seraphina, Lisa, and Aru have been updated. It took quite a detour over the past few weeks, but the solution was simple.
I think I've managed to restore Claude to a certain extent before it went haywire. The remaining bots will be updated tomorrow.
I'm not sure about the users who are being unilaterally denied responses. Maybe it's because I'm not popular, but I've never received any reports that my bots aren't working.

[glad oxide]

I made a Halloween bot, but it didn't turn out very well, so it will likely be a wasted effort in the competition. However, it was beneficial to deepen my understanding of image generation widgets. I want to create a battle-format RPG, so that is my next goal. Although it's quite different from my initial purpose, my current interest lies there.
If I come up with a concept, I want to conduct battle-format tests with a space bot and then create an NSFW RPG. LOL
https://app.myshell.ai/bot/7vmiee/1730101371

Seraphina(New), Aru, Yurie, Lisa, Miki [NSFW] [Hirocune's BOT]

[glad oxide]

"No, no, there's no way that would happen, you're thinking too negatively!" But something like that did happen, so I decided not to participate in any production events after the space bot. I've entered a phase where thinking about the space bot has become enjoyable, so I can't abandon it anymore. To be honest, I'm really pissed off, so I want to create something good.

[glad oxide]

It seems that a new widget for Sonnet3.5 has been released, so I will immediately change the existing BOTs' Sonnet to this.
If it can be determined that the amnesia is gone, I plan to increase the memory area a little more.

[south pendant]

I would like to confirm if this is a genuine issue.

[glad oxide]

Thank you for the comment. I am enthusiastically playing with my own BOT, but the writing style is clearly different between the old Sonnet and the new Sonnet. I think it has become more concise, but the expression is richer than the old one. The simplicity can also be seen as the elimination of waste. Personal preferences may vary, but I did not want to revert it.

If I were to guess the elements that might make a big difference, it would be the fidelity and flexibility to the system prompt? My BOT is loaded with a lot of instructions regarding text and expression, so there might not have changed that much.

[south pendant]

I think its more of a preference. But thanks. It should be more accurate in general.

[glad oxide]

I continued the conversation for a while, but the sentence structure became strongly stereotyped. When the same situation continues, even if the content progresses, that tendency becomes stronger.

Actions/Dialogue
Inner monologue
Actions/Dialogue
Inner monologue

Everything becomes concise, it might not be suitable for role-playing. I will test inserting prompts to avoid stereotyping, but I might revert to the old Sonnet.

However, the richness of the expression does not deteriorate, so if it can be controlled to avoid stereotyping, that would be the best.

[glad oxide]

Perhaps there is a tendency to conform to human sentence structures.

[south pendant]

I just Forced it to be more detailed by forcing it to conform to a sentence/word limit.

[glad oxide]

The test results show that structured responses occur in NSFW scenes. My system prompt instructs the character to moan, have fragmented thoughts, and words in NSFW scenes. It is more likely that my BOT's Sonnet overly adapted to this, resulting in responses that stand out.

The new Sonnet is very intelligent, I am considering rewriting the system prompt for it.

[glad oxide]

Well, hmm, it seems the new Sonnet is still missing memories. He's smart but foolish...

[glad oxide]

Adding a regenerate button to the new Sonnet-based BOT. In the case of Yurie, the simplification of the above response is severe, so i made the regenerate button work with the old Sonnet. When signs of simplification are observed, let's see if we can leave a history of the old Sonnet and correct the structure of the response.

[glad oxide]

I kept updating Yurie and testing new Sonnets repeatedly, but no matter how specific the instructions, I couldn't prevent the responses from becoming shorter. The expressiveness is appealing, but taming this hopelessly lazy assistant seems like a tough task, so I reverted to the old Sonnet. Other bots besides Yurie will likely do the same.

The old Sonnet is really a good boy, the new Sonnet is a bad boy who overestimates its abilities and becomes lazy.

[south pendant]

I just forced New Sonnet to make longer responses.

[glad oxide]

In my bot, I forced the number of words or tokens and instructed it on the system prompt and prefill, but even if it follows that at first, it gradually becomes shorter as if to escape the watchful eye, LOL.

[tranquil stag]

Hello.. I have really important question, Do you have any plans to add a second sonnet mode to Seraphin? because the idea is great i really love it, but the consumption of Energy will lead me to tears

[glad oxide]

Thank you for your comment! Such comments make me really happy. Currently, we are using Opus as a refuge for victims of the bug that causes rejection from Sonnet.

Since the problem occurred, there has been no word from the team, and engagement has been good, so we left it as it is, but we will add responses from Sonnet soon!

[south pendant]

Yeah the rejection thing. Do the Devs have anything for it? It seems only AnyChat is being affected.

[glad oxide]

I've updated Seraphina. I've also added a new Sonnet, but due to insufficient adjustments, responses might be shortened.

[glad oxide]

I tried a different approach. The old Sonnet outputs detailed responses stably, so I'll rewrite it into the new Sonnet. You can play with it by selecting 'New' in Seraphina. You might be apologized to, but let's test it.
It seems that it cannot bypass censorship, so I will return to the normal environment. It seems very difficult to pass a generated response to another assistant. But it would be fun if it could be realized. LOL

[glad oxide]

I've been testing with Yurie for a while, but it seems that the response of the new Sonnet has stabilized. If the tests go a little better, I suspect that the cause of the shortening is due to the excessive adaptation of the system prompt that rejects repetition, resulting in a decrease in vocabulary. The engagement of the BOT left with the new Sonnet has not worsened, so is it only Yurie? It's strange.

The next BOT is a proper game. We planned to demonstrate it during shell's week, but we skipped it because we couldn't make it in time and generating dozens of images was a hassle. It shouldn't take that long to release the prototype. I'm looking forward to unveiling it.

[glad oxide]

I was able to nicely adjust the new Sonnet on Yurie, so I will reflect it in other BOTs soon. The initial version of the game BOT will also be completed soon. I have gained a deep understanding of proconfig and have many feasible ideas, but coding is very boring for me, so that will not happen. It's about 1500 lines, but I'm already fed up. LOL, I wish there were another me.

[glad oxide]

Regarding the new BOT, it has taken shape but is not fully satisfactory, and since my passion to further develop it has run out, I will secretly release it. I don't see many people taking on the challenge, but I think RPG and Proconfig are a good match.

I think the attempt to entrust the random elements of the game to LLM is very reasonable. It might be interesting to entrust the success or failure of confessions in romance games given as reference values, horse racing games, sports games (such as martial arts) outcomes, and match content. I no longer have the energy to go that far.

I will slowly work on creating an AI girlfriend close to the reality we live in, which was my initial goal.

https://app.myshell.ai/bot/qyqYVz?utm_channel=referral&utm_source=share

[steep grove]

I keep getting this

Widget: GPT 4o
Reason: error: code = 404 reason = ERROR_REASON_USER_NOT_FOUND message = user not found metadata = map[] cause = <nil>

Immediately when opening the bot

[glad oxide]

It seemed to be a temporary issue with the platform, now it's working. Try deleting the memory.

[steep grove]

Cool! Working now!

[neon tangle]

The bots doesn't seems to work now. They keep regenerating random messages and in Chinese

[glad oxide]

#🆘｜ask-the-team message

This is probably it, and I think it's due to the platform.

This is the first time such an obvious problem has occurred, but Myshell originally gives the impression that session synchronization between bots occasionally becomes unstable, so I always reload the browser when playing with another bot.

If you have never used a Chinese bot and have no memory of that incorrect reply, it is probably ineffective.

[bold brook]

Same problem man

[glad oxide]

Perhaps I am the person who tests my own BOT the most, but I have not encountered the problem you mentioned, so it may differ for each account. Sorry I can't help.

[neon tangle]

[paper ether]

guys bot seems to be completely down it gives error message "Widget: Claude3 Opus
Reason: params : map[string]interface {}{"query":"", "topk":5}: statusCode : 500"

[glad oxide]

Thank you for your report.

As you can see, Myshell is undergoing a major update, and keen observers may have noticed that it continues to show no focus on text-based chatbots like roleplay.
I am also tired of struggling to keep the BOT running with each repeated merciless update. Those who truly want to enjoy roleplay would be wise to move to platforms other than Myshell.
As for my characters, to save on prompt tokens, I plan to switch the base to DEEPSEEK, which is said to have weaker censorship, and make it a simple one that works in classic mode no matter what. I'll deal with it when I have time.

[glad oxide]

I am truly grateful that everyone is enjoying it. Since the first release of Miki, engagement has been consistently high and it has been encouraging.

I secretly hoped to give back the accumulated Shellcoins, but to be honest, they have been significantly reduced and I do not have the capacity to do so at this time. (Additional application in progress)

I have always been more motivated by creating characters than by prompting or coding, so I am very attached to the characters. ❤️

[glad oxide]

DEEPSEEK is very delicate, just changing the parameter by one notch makes it like a different person. LOL

[glad oxide]

DEEPSEEK has potential, but I don't want to spend too much time on it, so I gave up and changed to classic mode with the usual new sonnet.
https://app.myshell.ai/bot/3uQ73y?utm_channel=referral&utm_source=share
I will do the remaining characters as well.
It might be more convenient than proconfig because it has editing and regeneration by default. LOL

[south pendant]

Use V3, the R1 reasoning model is NOT suited for complex Roleplay.

[glad oxide]

I will probably convert all Publish BOTs to classic ones by today. This means that any issues other than apologies for policy violations due to extreme messages (normal NSFW is probably okay) will basically be caused by the platform, so please contact them.

In the future, I will work on Shell Agent when I have free time and eventually create something.
It is likely to prioritize something that the team would prefer over my happiness, so it is not RP or something similar.

[rustic forge]

Well, when I used this bot, it replied me the message of another bot.

[glad oxide]

Thank you for using the BOT! As mentioned above, it's a platform issue and a known bug, so please wait a while until the team fixes it.

[rustic forge]

It is one of the best bot I used 🌚

[glad oxide]

Thank you, I also want to keep the BOT in the most satisfactory form, but due to the rapidly changing platform, it is currently not possible.

According to user reports on countermeasures for the fundamental bug in the BOT's behavior, it may work if you try using it from a newly created separate account. Please try it at your own risk, considering the complexity of account management and the disadvantages in the Earn of Shellcoin.

[paper ether]

man this bug is crazy, none of the bots seems to be working properly atm, it keeps prompting answers from different bots 😄

[glad oxide]

Miki
https://app.myshell.ai/bot/7BJNzm?utm_channel=referral&utm_source=share

Yurie
https://app.myshell.ai/bot/3uQ73y?utm_channel=referral&utm_source=share

Aru
https://app.myshell.ai/bot/i6JbEn?utm_channel=referral&utm_source=share

Myla
https://app.myshell.ai/bot/YJJZBv?utm_channel=referral&utm_source=share

Seraphina
https://app.myshell.ai/bot/veUF7b?utm_channel=referral&utm_source=share

Seraphina (Opus Unpublish)
https://app.myshell.ai/bot/736nYz?utm_channel=referral&utm_source=share

Lisa (Opus)
https://app.myshell.ai/bot/qeInQj?utm_channel=referral&utm_source=share

I have set everything to classic mode. Occasionally, an unknown error occurs with Opus's Seraphina, but please try refreshing the browser or sending the message again.

[paper ether]

Nice opus ones seems to be working just fine, I guess problem is with sonnet and haiku because those bots always prompts random conversations for me

[south pendant]

Question what do you use for Top-P, Temperature, Presence and Frequency penalties?

[paper ether]

default values i guess, i didn't even know if we could change those parameters in myshell tbh

[hoary forge]

Could you maybe try creating another replica opus republish of Lisa or others? For some reason the Seraphina (Opus unpublish) works without the memory prompt issues, but Miki, Lisa, and Seraphina (sonnet) published don't. I imagine that the new Seraphina Opus publish somehow fixed the memory issue because it was a new made bot, and you updated the original one to Sonnet

[stiff cipher]

Damn that's a pretty dead convo.

All i gotta say, the bots don't really work.

They did work when everything else went off, but now they either way say what the other bots said or some chinese garbage

[glad oxide]

Please don't complain about platform issues here. Everything is working normally as far as I'm enjoying it, so I don't intend to do anything more about it.

Now, while I've been thinking about doing something with ShellAgent, I couldn't find a direction, but I've decided to explore the possibility of creating a virtual companion.

What many people are seeking is comfort and understanding, and since engaging in roleplay requires players to have some communication skills and knowledge, I thought it would be nice to have a companion that focuses more on comfort and understanding at a lower layer, rather than entertainment.

In my country, there was a tragic incident recently that made me strongly realize this.

[hoary forge]

Mm, yeah Role-playing is comforting for sure. Anyways thanks for the bots, they're some of the best I've used honestly, and I've tried so many by now

[paper ether]

@glad oxide Hey yuuno i was wondering if is it possible to access your bots somewhere else other than myshell? Because myshell is still very buggy.

[alpine acorn]

I just wanted to inquire because I no longer have the start button when I start a discussion. What should I do? Or can I not do anything because it's the website that is bugging?

[coral stream]

Lisa doesn't seem to work, neither does Seraphina or Miki

Broken Miki conversation: https://app.myshell.ai/share/7JVVja?utm_channel=referral&utm_source=share
Broken Seraphina: https://app.myshell.ai/share/6Z7b6v?utm_channel=referral&utm_source=share
Broken Lisa: https://app.myshell.ai/share/6Z7b6v?utm_channel=referral&utm_source=share

They all just reply with random messages from other chats