Blog
Essays, field notes, legal-tech analysis, and how FlowCounsel™ builds and delivers.
AI & Technology
AI tools have compressed the cost of producing code, demos, and polished interfaces. They have not compressed the work of building systems that hold up in production.
April 18, 2026
AI & Technology
AI-enabled cyber threat research and emerging agent-worm research point to the same architecture problem legal AI has to solve: models become dangerous when connected to tools, memory, and execution loops without control state.
June 3, 2026
AI & Technology
Agent runtimes need isolation, tool boundaries, and monitoring. Legal AI needs something more: record truth, provenance, verifier state, review state, and approval control.
May 28, 2026
AI & Technology
Thomson Reuters and Anthropic just strengthened the lawyer's assistant category in a serious way. The larger fight still sits underneath intake, matter state, review, and the record legal work runs on.
May 21, 2026
AI & Technology
The legal AI market keeps arguing about future regulation while ignoring a more immediate issue: too many systems collect, retain, route, and expose sensitive legal-intake and matter data with weak discipline.
May 14, 2026
AI & Technology
Microsoft's new Legal Agent in Word is a serious product and a serious signal. It strengthens one category of legal AI: the lawyer's document tool. The next category is the firm's system.
May 5, 2026
AI & Technology
The most consequential error in AI discourse is mislocating where the cognition lives. LLMs can produce output that looks like reasoning, but legal AI products have to be designed around what the systems actually are.
April 18, 2026
AI & Technology
A Singapore court decision on algorithmic crypto trading is one of the clearest judicial frames for agentic-system risk: when software acts without a human checkpoint, responsibility traces back to how the system was designed.
April 8, 2026
AI & Technology
Anthropic's Mythos Preview and Project Glasswing are not just cybersecurity stories. They are a warning that legal AI governance can no longer focus only on hallucinations and output review. System access is now part of the risk model.
April 8, 2026