We're building AI conversations indistinguishable from real humans. Our mission: pass the complete Turing test.
Current AI roleplay models seem fine for short chats. But extend the conversation — and they forget, lose emotional coherence, and wander aimlessly.
The root cause? Training data. Models learn from scraped forums, AI-generated filler, and shallow scripts — none of which capture how real people connect over time.
Human-written, structurally rich dialogue data. No scraped content, no AI-generated filler. Every line is crafted by real writers who understand how people actually talk.
We build complete character arcs with memory, emotional depth, and narrative purpose — training data that teaches models to be genuinely human.
The personalities our writers bring to life
Hi, I'm the on-call Cao Cao. Welcome to Singularity Gate Research.
“What kind of script is this? What kind of character design? Why am I dressed like this?”
“Because being on-call is very AI. Wearing the Cao Cao costume makes you feel more human.”
“Are you kidding me?!”
From concept to reality
Website development and core feature implementation. Building the foundation for the writer portal.
Public announcement on Bilibili. Demo release and first wave of writer recruitment.
Character finalization, team assembly, and the beginning of structured data collection.
Data collection milestone reached. Begin model fine-tuning with our human-crafted dialogue dataset.
Model evaluation, iteration, and beyond. The Turing test awaits.
“If I had asked people what they wanted, they would have said faster horses.”— Henry Ford
“We will complete the Turing test.”— Singularity Gate
Choose your path to get started