Writing
Notes
Short pieces. Where an essay started on LinkedIn, the page says so and links out.
- AI securityHow to keep authority over agents that can call tools and touch systems of record.
- Agentic AIWhat changes when software does not only answer a person, but takes an action on its own.
- AI governanceDeclare, record, govern, and measure the authority an AI agent is allowed to hold.
- AI-native transformationHow an organization changes the way it works once agents are in the operating model.
- What the Trust-Autonomy Gap measures
What the Trust-Autonomy Gap measures
The gap is the distance between what an agent can do and what the organization can see, justify, and revoke. You close it one agent at a time.
An introduction to Google Antigravity
Google's agentic development platform marks how building software is changing. This note says what the shift is, in my words, and points at the original essay.
When a creator runs their own model
A large creator training a model at home is a signal for enterprises, not a gadget story. The question is who holds the weights, the data, and the off switch.
The AI-native change agent
Someone has to carry the new way of working into the teams that will live with it. The role is about adoption with controls, not a slide.
Distributing agency across agents
Multi-agent tools split a job across specialists. That only helps if each agent's authority is still explicit and reversible.
From AI hype to a result you can defend
A trusted advisor in this work is someone who can say what changed in the operating model, not only what the demo did.
Signals, not noise, in an AI strategy
Most of the noise is a tool announcement. The signal is a decision about authority, data, and who can stop the system.
Reasoning models, and what they change
Models that spend more steps before they answer change review, not just quality. A longer chain of action is a longer chain of authority.