Last updated: February 2026 · Nairobi
Deep in the Bluedot technical AI safety curriculum. The focus right now is on interpretability specifically how to extract decision-relevant features from transformer internals without drowning in noise.
Also leading partnerships at GDG Pwani, which means a lot of coordination work for the upcoming Build with AI season.
The politics of AI governance. Who gets to decide what "safe" means, and whether the current landscape is trending toward pluralism or priesthood. (I wrote about this recently.)
Also: the gap between alignment research and deployment reality. The field has a theory-practice problem that nobody wants to name directly.
Re-reading Seeing Like a State (Scott) — keeps being relevant. Also working through the Anthropic interpretability papers to trace how the field's assumptions evolved.
A small tool for visualizing attention patterns in a way that's actually useful for debugging. Early stages. May never ship.