Deterministic where possible
A useful boundary for AI systems is to keep judgement in the model and move invariants, validation, and repeatable transformations into code.
Writing / Notes
Short technical notes
Working notes rather than a polished publication schedule: mechanisms that seem useful, things learned from experiments, and ideas worth making explicit enough to revisit or falsify later.
Evaluation is system design
Evaluation becomes more useful when it is designed with the workflow, not added after the model already appears to work.
Convergence over capability
The next framework, agent, or tool is only useful when it resolves a concrete blockage better than the mechanism already in hand.