6 min read
Require Evidence for AI-Generated Code
Use explicit acceptance criteria, test results, risk-based review and production feedback before accepting AI-generated changes.
Recent Insights Page 4
7 min read
Why Single Agents Often Beat Multi-Agent Systems
Why many apparent multi-agent gains are really test-time compute gains, and when extra agents are still worth the complexity.
3 min read
We Had This Argument in 1995: Restricting Cybersecurity AI Is a Losing Game
Gated cyber models may buy defenders time, but history suggests the advantage will not last.
10 min read
The Automation Avant-Garde
When machines master the ordinary, humans are freed to pursue the extraordinary. From photography to AI music to vibe coding, the same pattern repeats.
6 min read
We Gave AI the Keys to Production. Now What?
AI coding tools and autonomous agents are shipping faster than the guardrails meant to govern them. Here is where the risks are and what thoughtful adoption looks like.
4 min read
Restricted Frontier AI Needs Public Access Rules
If frontier AI access is restricted for security reasons, the criteria should be narrow, reviewable and available to public-interest institutions.
8 min read
What We're Getting Wrong About AI
The AI debate keeps collapsing into hype, panic, and lazy binaries. The real issues are truth, work, power, and governance.
16 min read
AI agents in monorepos: what to configure differently from a single-product repo
A practical guide to running AI coding agents safely and efficiently in a monorepo, including the wins, the configuration changes, and the common failure modes.