Justin Reock

Justin Reock is Deputy CTO at DX, where he works on developer productivity and the practical impact of AI-assisted software development. His central concern is what happens after code gets easier to produce: whether teams deliver useful changes faster, how confidently they can operate those changes, and where the saved time goes.

From enterprise infrastructure to developer productivity

Reock’s earlier work centered on enterprise software and open-source infrastructure. He worked on messaging middleware at EarthLink and OpenLogic, developing an interest in how separate applications communicate reliably. His technical writing made that machinery approachable: his Apache Camel examples connected integration patterns to working implementations, while his Prometheus and Grafana guidance explained how to configure alerts that help teams act on monitoring data.

As chief architect at Perforce, Reock advocated for open source as a foundation for enterprise innovation. His contribution to the 2019 Open Source Support Report placed machine learning, data infrastructure, and Kubernetes within a larger shift toward software developed collaboratively across organizations. His focus was how enterprises could use and support that community software.

At Gradle, his work moved toward developer productivity engineering: improving the build and test systems that determine how quickly engineers receive feedback. As Field CTO in 2023, he developed guidance for build-system migrations that made observability part of the transition. Build scans could reveal which projects had migrated, explain unexpected dependency selections, and give colleagues a shared view of a failure without rounds of copied logs and follow-up questions. Faster feedback meant less time spent reconstructing what had happened, as well as faster builds.

Reock subsequently became Head of Developer Relations at Cortex. He and Lauren Craigie co-hosted Prefrontal, exploring internal developer portals, platform engineering, and the organizational conditions behind productive engineering work. His focus expanded from individual infrastructure components to the systems through which teams find information, coordinate responsibilities, and deliver software.

When he joined DX as Deputy CTO, Reock brought that experience into research and education about engineering productivity, connecting the role to his longstanding interest in DORA, SPACE, and developer-experience measurement. AI sharpened a question running through that work: how much does improving one activity help when the surrounding workflow remains constrained?

Measuring AI’s effect on delivery

Reock’s work at DX separates AI adoption from its consequences for engineering teams. Tool usage and reported time savings matter, but neither explains on its own whether an organization delivers more value or incurs additional costs.

• Measure utilization, impact, and cost. Reock helped introduce DX’s collaboratively developed AI Measurement Framework, which separates tool use from engineering outcomes and expenditure. It combines product telemetry, developer surveys, and targeted experience sampling to examine delivery, quality, and developers’ confidence in their work. A central question is what the organization gains from the time engineers report saving.

• Find the actual bottleneck. Reock argues for tracing work from an idea through to customer value before deciding where AI should intervene. His analysis of DX’s longitudinal data reported a 7.76% increase in pull-request throughput alongside a much larger increase in AI usage in the studied sample. He treats such gains as meaningful while challenging expectations of automatic multiples in productivity. Planning, alignment, review, and handoffs can still limit delivery even when implementation accelerates.

• Distinguish understandable code from trustworthy changes. In his Q2 2026 assessment, Reock highlighted a divergence between developers’ ability to understand their codebases and their confidence that changes would avoid production failures. He also examined larger pull requests, additional review friction, rising AI expenditure, and reported time savings that had not increased the share of time spent building new features. These findings explain why he considers velocity, quality, developer experience, and spending together: improvement in one measure can coexist with new pressure elsewhere.

Building reliable human-agent workflows

Reock pairs measurement with practical guidance for making AI useful in complex engineering environments. His recommendations address both how engineers direct models and the conditions in which people and agents work.

• Use structured prompting for complex systems. His advanced prompting guidance moves beyond conversational requests toward explicit constraints and validation. Graph-based prompting exposes dependencies between rules; controlled validation loops check results; comparisons between separate implementations provide an additional check for critical work; and diff-only refactoring limits unnecessary changes. Each technique gives teams a way to test or constrain a model’s plausible answer.

• Improve agent experience through better engineering foundations. Reock extends developer-experience principles to coding agents: current documentation supplies usable context, modular code makes tasks more tractable, and fast CI shortens the feedback loop for both people and machines. His systems approach to AI adoption also emphasizes education and psychological safety, so engineers have time and permission to experiment and learn. He describes DX experimenting with surveys of agents after completed tasks to identify the friction they encounter.