Writing about agent engineering, AI, and legal tech. https://blog.junsuzuki.xyz Karpathy's LLM Wiki applied to a codebase, built up one session at a time. The same shape Meta reached with heavier machinery. Map-as-you-go. https://blog.junsuzuki.xyz/blog/load-less-remember-more https://blog.junsuzuki.xyz/blog/load-less-remember-more Mon, 18 May 2026 00:00:00 GMT Paper co-written for the CNC Lab on AI and creation. AI originality isn't a slider, it's a narrow zone between conformity and collapse. https://blog.junsuzuki.xyz/blog/singularity-controlled-divergence https://blog.junsuzuki.xyz/blog/singularity-controlled-divergence Wed, 29 Apr 2026 00:00:00 GMT Anthropic packs domain knowledge into Claude Code skills. We built a separate knowledge layer. Turns out it's the same architecture. https://blog.junsuzuki.xyz/blog/anthropic-skills-playbook-custom-knowledge https://blog.junsuzuki.xyz/blog/anthropic-skills-playbook-custom-knowledge Sat, 04 Apr 2026 00:00:00 GMT CLAUDE.md doesn't scale past a few hundred lines. Here's how we built a repo-scoped knowledge layer that gives Claude Code the right context at the right time. https://blog.junsuzuki.xyz/blog/beyond-claude-md-repo-scoped-knowledge-layer https://blog.junsuzuki.xyz/blog/beyond-claude-md-repo-scoped-knowledge-layer Wed, 18 Mar 2026 00:00:00 GMT