All writing

The full archive, newest posts first.

When AI Writes Most of the Code, Auditability Becomes the Product

Anthropic’s claim that Claude authored more than 80% of merged production code is not mainly a model-quality story. It is a software governance story about reviewability, provenance, rollback, and how enterprise engineering changes when generation stops being the bottleneck.

June 8, 2026

aienterprise-aisoftware-engineeringai-agentsgovernancecode-generation

Enterprise AI’s Hard Problem Is Control, Not Capability

The real enterprise AI bottleneck is no longer raw model capability. It is control: permissions, auditability, approvals, observability, and cost discipline around systems that are already capable enough to create operational risk.

June 8, 2026

enterprise-aigovernanceai-agentsarchitectureoperations

California Is Turning AI Procurement Into an Architecture Question

California’s new AI executive order matters because it pushes procurement to demand operational evidence from vendors, turning governance from policy theater into a real architecture and delivery issue.

April 5, 2026

aienterprise-aiai-governanceai-procurementai-architecturepublic-sector

Google’s New Gemini Inference Tiers Matter More Than Another Model Launch

Google’s Flex and Priority tiers matter because they turn mixed-criticality AI workloads into a first-class production design choice instead of forcing enterprises to fake it in their own architecture.

April 4, 2026

aienterprise-aigooglegeminiai-operationsagent-architecture

Google Didn’t Just Launch Gemma 4. It Removed One of Enterprise AI’s Biggest Excuses.

Gemma 4 matters less because it is just another strong model, and more because Apache 2.0 makes it a more credible enterprise deployment option.

April 3, 2026

aienterprise-aiproductiongooglellmgemma