SMFIELD / NOTES

Notes from the work

Ideas for the room where decisions happen.

Short essays about AI, security, infrastructure, and leadership—written for operators who need the idea to keep working after the meeting ends.

Explore
AI governance

Govern the decision, not just the model

The practical unit of AI risk is not the model in isolation. It is the decision a person or system makes because of the model’s output. That distinction changes the questions an executive team should ask.

Start with consequence. If the output can deny access, move money, change care, expose data, or trigger a regulatory obligation, the workflow deserves stronger review, logging, and human authority. A low-consequence drafting assistant and a high-consequence eligibility engine should not share one generic “AI policy.”

Then define the exit. Who can override the system? What evidence will they see? Can the organization reconstruct why an answer was accepted? Governance becomes real when those answers exist before production.

Read the full field note →
Incident leadership

The first status update should reduce uncertainty

During an incident, activity is easy to produce. Shared understanding is harder. The first useful update does not need a perfect cause; it needs a stable frame.

State what users experience, what is confirmed, what remains unknown, and the next evidence checkpoint. Keep hypotheses labeled as hypotheses. Assign one owner to the timeline and another to the technical work so investigation does not erase communication.

The best update is not the longest. It gives the room enough truth to make the next decision without pretending the investigation is finished.

Read the full field note →
Team craft

“Hands-on” is a method, not a hero story

Leading from the trenches should not mean becoming the bottleneck with the fanciest keyboard. It means staying close enough to the work to remove fiction from the plan.

A hands-on leader can see when the interface is misleading, the runbook is stale, the recovery window is imaginary, or a team is carrying risk it cannot name upward. The response is not to seize every task. It is to clarify the system, improve the conditions, and return ownership stronger than it was.

Read the full field note →
AI infrastructure

Why serious AI infrastructure is a governance decision

AI infrastructure is never only a compute choice. In genomics and medical research, the architecture determines where sensitive data travels, which controls remain enforceable, how work can be audited, and whether an intensive model can run without turning security into an afterthought.

This field note uses Sergio’s publicly documented work on hybrid cloud and NVIDIA infrastructure for Project ZenQ-AI as a practical frame for the tradeoffs.

Read the full field note →