Craig McLuckie on Culture as a Team's Operating System in the AI Era

McLuckie’s core point: AI adds a new maturity axis. Without process and culture changes, teams get more code, larger PRs, and slower feature throughput. Open source maintainers are already seeing “AI-slop” PR floods that break norms like “good first issues,” which used to onboard contributors but now attract low-quality automated fixes. Expect similar dynamics inside companies if incentives and gates don’t change.

The risk profile is shifting. AI increases output but also defect and exploit risk. Seniors become overburdened reviewers and de facto managers of stochastic agents. Organizations that measure lines-of-code or “AI usage” will accidentally reward bloat and push review debt downstream, eroding morale and quality. Claims of massive productivity are mostly anecdotal and fragile without evaluation harnesses, security scans, and small-change discipline.

Culture is your operating system. Make it explicit, tie promotions and decisions to it, and evolve it deliberately (hypocrisy kills it). In the AI era, anchors should include ownership of AI-generated code, small atomic PRs, reproducibility, and safety/reliability as first-class goals. Code is trending toward an intermediate artifact; value shifts to specification, review, and risk management. Hiring, leveling, and promotion criteria must follow.

Practical shifts: institute AI governance (tooling, data boundaries, provenance in commits), PR size limits and review SLOs, mandatory tests and security checks for AI-originated code, and agent evaluation frameworks with task-level metrics. For open source, gate “good first issue” work via assignment, templates, and test proofs to curb spam. Rethink junior development: apprentice via evaluation writing, guardrails, integration tests, and reliability rotations since traditional “low-level tasks” are automated. Treat agents as stochastic systems; design via experiment/eval loops, not big upfront design.