Insights
Practical notes on building reliable, governable systems that remain understandable as they grow.
Analytics Delivery · 31 August 2026
A controlled study of modular in-context-learning text-to-SQL systems suggests that execution-feedback refinement should be tested before model upgrades. The practical lesson is not to copy a benchmark stack wholesale, but to evaluate models and pipeline modules as one governed system under accuracy
Read articleAI Governance · 24 August 2026
A practical framework for deciding when to compile source-supported claims at ingest time instead of repeatedly interpreting raw text at query time—and for operating the resulting semantic index with provenance, maintenance, migration, and cost controls.
Read articleLeadership · 21 March 2026
A practical architecture rule for keeping governed AI workflows understandable, testable, and safe to change.
Read articleA controlled study of modular in-context-learning text-to-SQL systems suggests that execution-feedback refinement should be tested before model upgrades. The practical lesson is not to copy a benchmark stack wholesale, but to evaluate models and pipeline modules as one governed system under accuracy
A practical framework for deciding when to compile source-supported claims at ingest time instead of repeatedly interpreting raw text at query time—and for operating the resulting semantic index with provenance, maintenance, migration, and cost controls.
A practical architecture rule for keeping governed AI workflows understandable, testable, and safe to change.