Own Your AI (Or Your AI Owns You)
Lawyers obsess over model choice. It's the least durable decision in the stack, and the wrong one to commit to. Build the system, not the dependency.
Read moreArchive
All 9 posts, newest first
2026
Map, Not Manual: How I Structure Every AI Project
LLMs are trained to please. For serious work, that's a liability. The instruction file is how you fix it — once, at project level.
Read moreEncoding Legal Expertise into AI Agent Skills
Output quality tracks skill quality. This is how I build agent skills that hold up in M&A deal review — and the patterns that make any legal workflow encodeable.
Read moreWhat Your Agent Is Doing When You're Not Watching
AI agents don't fail loudly — they fail invisibly. What observability means in practice, how to set it up, and why it matters for builders and users alike.
Read moreWhat Agents Need (And What They Actually Are)
An agent is four components: LLM, tools, loop, stopping condition. Simple. What it actually needs to work is a longer list, and most (legal) tech wasn't designed to provide it.
Read moreThe Vibe Coding Trap: Why 'Just Ask AI' Isn't Professional Development
Vibe coding is brilliant for prototypes, but production legal tools require engineering discipline. Here's why the difference matters, and what changes when you build with AI assistance.
Read moreThe Claude Legal Plugin Panic: Confusing the Foundation Layer for Competition
Anthropic's legal plugin announcement triggered panic in legal tech stocks. But it's not a disruption, it's a consolidation of existing capabilities.
Read moreThe Structured Data Mindset That Unlocks Legal Automation
Legal work is data work. Once you start seeing information as structured data instead of documents, you unlock automation, better knowledge management, and effective AI integration.
Read more