Practical AI workflows and developer tools from Tokyo

Hi everyone, I'm Yusuke Hayashi, a Tokyo-based software engineer and AI adoption consultant at haya Inc.

I help teams turn AI ideas into working products: redesigning workflows, testing useful pilots, and taking them into production. I also keep building tools myself:

• BlobDive: bounded, read-only inspection of nested software artifacts, written in Rust.
• MCP Erabi: a Japanese interface for comparing MCP server connection requirements.
• Kata Prompt: a local-first workbench for assembling and reusing Japanese prompts.

I'd love to compare notes with people building agent tools, MCP integrations, and AI workflows that teams actually use. Happy to contribute what I learn about reliability, evaluation, and rollout.

What is one practical bottleneck in your AI workflow right now? Especially interested in artifact inspection and Japanese-language workflows.

Projects: github.com/yhay81
About my work: yusuke-hayashi.com