Rashomon strikes in the Court of Appeal

After 150+ hours building data.zeeker.sg, I attended SMU's SOLID database launch to see institutional legal data infrastructure in action. I applied my decision framework and made my choice: I'm continuing. Here's what I learned about when solo builders should persist alongside institutions.

The complete technical stack behind Singapore's first public legal news API: zeeker CLI standardizes data collection, Datasette serves SQLite straight from S3, and canned queries make legal research accessible without SQL. Built solo, runs on $6-12/month, designed to scale from 1 to many sources.

This Hacktoberfest, I'm not contributing code—I'm maintaining my own. Two years after redlines went viral, I'm adapting it for AI agents. The lesson: tools that don't work with AI will fade. Open source needs to evolve with how people actually work.

A new term is making headlines: "workslop"—AI-generated work that looks polished but lacks substance. Research found 40% of workers encountered it last month, costing $186 per employee in cleanup time. For resource-constrained lawyers, you can't spend time fixing what's irretrievably broken.

Singapore's High Court just handed down what might be its first formal ruling on AI-generated fake cases in legal practice (2025 SGHCR 33). For resource-constrained practices, this case validates a hard truth - you might be better off avoiding AI entirely.