I build architecture around ambiguity — then write about what I find inside it. My work sits at the intersection of AI governance, organizational design, and trust infrastructure. I've led AI strategy at Microsoft, built ventures where wrong answers are expensive, and spent most of my career asking one question: what makes a structure trustworthy enough that people willingly place part of their finite life inside it?
Currently Principal Technical PMM at Atlassian (Rovo), co-building WriteByte, and writing frameworks — Directional Delegation, harness debt, the Representation Layer — that treat AI deployment as a design problem, not a speed problem.
Most of what I publish starts as a hunch in conversation — overheard phrases, a tension I can't let go of, a framework that suddenly connects two domains that shouldn't talk to each other. I work with AI the way I work with good collaborators: I bring the wedge, the structural instinct, the "this matters and here's why." The prose gets shaped in the back-and-forth. I care about provenance not as compliance but as honesty — if a sentence emerged from meld, it should say so.
Tensions I keep returning to: innovation vs. ethics, speed vs. governance, prevention vs. profit, founder leverage vs. institutional leverage. These aren't contradictions to resolve — they're load-bearing tensions to design around.
Recurring obsessions: Why organizations stay stuck when individuals get faster. How trust degrades at the seams between human and machine judgment. What provenance actually means when AI writes half the sentence. Whether restoration is a more honest frame than revolution.