What we think about,
in public.
Technical and commercial writing on building AI systems that reach production. No predictions, no hype cycles — the problems we actually run into and how we handle them.
- AI Security7 min
Prompt injection is an authorization problem
Treating injection as a prompt-engineering issue guarantees you will not fix it. The durable controls sit in the tool surface and the permission model.
- AI Modernization6 min
Why legacy systems block enterprise AI
The blocker is rarely the model. It is that the data an AI system needs is trapped behind an interface designed for a human with a keyboard.
- AI Governance7 min
How to govern AI agents without stopping them
Governance that says no to everything gets routed around. The workable version is an inventory, a risk tier, and an approval gate proportional to consequence.
- AI Automation5 min
Which workflows are worth automating first
The best first automation is boring, high-volume, and has an unambiguous definition of a correct outcome. Ambition is what makes second projects fail.
- Enterprise AI7 min
How to build an internal company AI that people trust
Retrieval quality, permission fidelity and honest refusal decide adoption. Model choice is far less important than any of the three.
- Emerging Technology6 min
MCP, tool calling, and what agent interoperability changes
Standardizing how agents reach tools moves the hard problem from integration plumbing to permissions and provenance. That is a better problem to have.
Where our writing focuses.
Agents, automation, enterprise AI, modernization, governance, security and the emerging technology around them — MCP, agent interoperability, computer-use AI, voice AI, observability and private deployment.