At Rootcomputer, our focus has always been on the "inner lives" of artificial systems—understanding how they reason, how they are aligned, and how we can ensure they remain safe companions to human thought. Today, we are sharing a look at a tool we’ve been developing internally to streamline that process: Haiku Studio.
Haiku Studio is a standalone desktop application designed for the end-to-end creation of small language models (SLMs). While much of the industry focuses on massive, opaque clusters, we believe there is immense value in "local-first" development where the researcher has total control over the training pipeline.
The application allows you to move through the three critical stages of model development within a single graphical interface:
Everything runs locally on your own hardware—whether that is a CUDA-capable NVIDIA GPU or Apple Silicon—ensuring that your data and your model’s "personality" stay entirely under your roof.
We are currently in the early beta phase. The core engine is functional, but we are still in the process of debugging, refining the UI, and hardening the training stability. It isn't a "final" product yet, but rather an exploratory research tool that we are opening up for feedback.
Our roadmap for Haiku Studio includes:We aren't interested in just building another "bot maker." We want to provide a dedicated laboratory for those who care about the philosophy and safety of the models they create.
If you’re interested in the technical architecture or want to see the progress of the beta, you can find the repository here: