# Greg Wilson (third-bit.com)

Greg Wilson writes at third-bit.com on empirical software engineering and research methods. He founded Software Carpentry, the training project that taught scientists basic computing and reproducible practice, and edited the "Beautiful Code" and "The Architecture of Open Source Applications" collections. The recurring stance in his AI-era writing is that software engineering never learned from the human sciences how to study its own practices, so most claims about tools — AI-assisted coding, but also agile and TDD — rest on study designs that a psychology or medicine reviewer would reject on sight.

The writing is compact and reference-backed. Rather than argue that AI coding is good or bad, he argues about how you'd know, naming the specific methodological failure behind each common measurement and citing the study that illustrates it. He runs a one-day workshop and a one-hour talk on the underlying research methods.

## Topics

- Measurement and study design for software-engineering claims
- Research-methods failures: Goodhart's Law, Hawthorne and novelty effects, selection bias, internal validity, systems thinking
- Teaching computing and reproducible practice (Software Carpentry)
- Open-source architecture and code quality

## Ingested articles

- [[twelve-ways-wrong-ai-coding]] — his 2026 catalog of twelve measurement errors in studies of whether AI coding tools work, each mapped to a named research-methods failure

Related concept extracted: [[measuring-ai-coding-productivity]].