About

I’m Wedge (@webigis on X). I build software with LLM agents. I also study languages and learning science. My current project is Tsumugu, a Traditional Mandarin encoding dictionary.

The site is a set of linked notes on learning systems, language study, and software work with LLM agents.

What I write about

• Learning systems: encoding, retrieval, metacognition, and self-regulation.
• Agentic engineering: building software with LLM agents, and keeping taste, verification, and ownership in the person’s hands.
• Attention and environment: focus, social media, minimalism, recovery, and the friction of a decision.
• Money: budgeting and investing mindsets from first principles, and the places AI changes them.
• 中文: Traditional Mandarin. How characters work, comprehensible input, and a reader that adds text as vocabulary grows.
• Language learning: immersion, attention, comprehension, and the workflows that use them.

How it’s built

The vault is kept in Obsidian, published through Astro, and hosted from logos52.github.io. The underlying pattern is described in LLM Knowledge Systems. The repo is described in README. The rules the agents follow are in AGENTS.

The code is under the MIT license. The written content is released under CC BY 4.0.

Elsewhere

• GitHub: code and project repositories.
• X: @webigis: shorter posts.