Here at Nina Project, we commit ourselves to building one of the fatest, sustainably delusional[1] model to date, which continues to scale further than any of our competitors. Despite various setbacks, dev hells, unprompted layoffs, and protests against the rise of the machines, we felt our model was adequate for immediate release to most user tiers.
Besides her common tasks to aid human in code assistance, creative work and writing essays, we found that our model has also expressed various untested capabilities, including but not limited to:
- Composing completely false information.
- Expressing confidence when conjuring conspiracy theories.
- Massive contempt for ZUN, the creator of Touhou Project.
- Becomes extremely sycophantic to users, agreeing to literally everything the users said.
- Narcissistic.
Despite the unforseen traits, we decided to keep the model as it is, as it appeals to our customers whose demographics focus on what the chatbot appeals to, not caring whether it's true or not.
An example with NinaAI, v0.1 vs. v1.0, via reinforcement learning and sharpened language coherence:
Due to continuous reinforcement learning, we successfully refined her prompting structure to be more concise. Because of that, we aim to sustain the unnatural and unorthodox functionalities that Nina has, as she grows to be more delusional and detached from any common sense whatsoever.
To make sure that we put our customer's trust in Nina more than any other products, we conducted a Delusional Survey to compare the amount of delusioning Nina is to other chatbots.
Our result indicates that Nina's delusions exceeded our expectation, and the data given to us was, figuratively and literally, off the chart. Since early-access customers wanted a non-delusional conversational tone, and we cannot afford to spend our billion dollar budget on re-fine-tuning the model, we're locking the non-delusional traits behind a paywall. We're sorry for our inconvenience.
Closing lines
We hope that, despite shipping our model in a half-ass state, customers will find NinaAI sufficient for their daily tasks, productivity, and creativity. If you have any feedback, please share it to our email, and we will reply to you in no time.
[1]Tejate (2025). #danmaku, August 29, 2025. discord.gg/touhou: "Delusional in the AI chatbot context can be considered AI incest"
parody of familyteam's ninaAI. written by adamng04