Simon Willison's Weblog

The most widely linked secondary analysis of StrongDM’s dark factory approach. Published February 7, 2026 — the day after StrongDM’s official post.

What Willison Adds

Willison goes beyond the StrongDM blog post to analyze:

• The unit test gaming problem — why the test isolation approach is the key architectural innovation, not the agent itself
• Digital twin feasibility — skeptical that DTUs are as easy to build as StrongDM suggests, but acknowledges the conceptual breakthrough
• The cost model — walks through the $1,000/day token math and what it implies for adoption
• Comparison to other approaches — how StrongDM’s holdout scenario model compares to other testing strategies

Why It Matters

Willison is one of the most trusted technical voices on LLM capabilities. His endorsement of the DTU concept as architecturally significant carries weight. His skepticism about reproducibility is a useful counterpoint to StrongDM’s confident framing.