Starpower .Technology
// grab the star. throw it. talk to it.
-
01
ACTIVE
Groupchat-native language models
Training models for many humans and many agents in one room: turn-taking, thread tracking, disagreement, collaboration, and multi-perspective reasoning under pressure.
-
02
ACTIVE
Complex reasoning from human experts
Building datasets with real researchers and domain experts — capturing how they actually think, then training models on those reasoning patterns.
-
03
ACTIVE
Small, powerful models
4B–100B parameters. We start small and train for efficiency: polished datasets, stronger reasoning loops, and models that use information to solve problems instead of only memorizing scale.
-
04
R&D
Open weights & new architectures
Open-weight releases plus ground-up architecture research: language-native tokenization, hierarchical meaning units, and SuperWVY / Savvy systems in active development.
-
05
SHIPPING
Autonomous frameworks
Starpower Autonomy — frameworks for agents that run their own research loops: YouTube Researcher, Autonomous Researcher, and beyond.
-
06
COMING SOON
WVY for iPhone
The multi-agent reasoning room in your pocket. Same collaboration loop, same agents, native on iOS.
Create SFT datasets right in the browser. A chat-style editor for user / assistant / system turns, one-click JSONL export, cloud save, and public sharing — the same training format we build WVY on.
Open Studio →The multi-agent collaboration room, native on iOS. Spin up agents with different roles and perspectives, then let them reason with you and with each other in real time.
slot 01
slot 02
The multi-agent collaboration room running on WVY. Humans and agents share one space, challenge each other's assumptions, split work in parallel, and bring multiple perspectives to complex reasoning.
Enter wvy.world →