Announcement · September 30, 2026

Hatsunama Publishes Independent Artificial Intelligence Research and Open-Source AI Systems Work

Hatsunama is an independent artificial intelligence researcher and AI systems developer whose work spans AI models, AI agents, agent harnesses, local AI, tool use, computer-use systems, AI evaluation, and applied AI engineering.

Hatsunama has published a public research index connecting model experiments on Hugging Face, open-source software on GitHub, and technical explanations of how models operate inside larger systems. Readers can inspect the linked work directly.

Models and software

Public model work includes a Qwen3 tool-calling LoRA and a repaired Qwen3 checkpoint. The LoRA card reports bounded tool-call tests and a remaining failure case. These tests assess model output in that setup; they do not establish that a complete agent finished real-world tasks.

Public software includes Caption Studio, a local captioning application, and Caption Action, an application for live captions and translation.

Evaluating the complete system

A model can propose an action, but surrounding software must select and execute tools, observe results, recover from errors, and verify the requested outcome. Hatsunama's research note on tool calls and task completion explains why plausible model output and verified task completion are different measurements.

The AI work page links specific artifacts. The research page describes the model-to-agent system perspective. Additional public work is available on GitHub and Hugging Face.

About Hatsunama

Hatsunama publishes open-source software, model experiments, data, and technical explanations about AI systems and applied AI engineering.