Engineering Leadership Insights
What I actually think about leading engineering organizations through the AI shift, written from production practice rather than from a conference stage. Each piece cites its sources and dates them, so you can see when the evidence was checked.
Judgment Is the Whole Job
When a team says most of its code is AI-written, that number measures keystrokes, not engineering. Producing code got cheap. Judgment about code got scarce, and the two statements people treat as opposites turn out to be the same statement.
Cites METR on long-horizon reliability, GitClear on rising duplication, and the Stack Overflow 2025 trust data.
Read the insight → InsightThe Architecture Review Bottleneck
When implementation goes from days to hours, the review queue backs up. The instinct is to speed review up. That is how you generate technical debt at machine speed and call it velocity. We made review slower and more expensive on purpose.
Three concrete changes, plus DORA on throughput arriving alongside instability and the review load practitioners actually report.
Read the insight → InsightWorking With Agents, Not Prompts
Prompt engineering is table stakes. Today's agents plan, edit across files, run tests, and open pull requests, and most teams still drive them like autocomplete. Lead with a thesis, force the full context, and expect the agent to push back.
The three practices that changed how my teams direct agents, anchored in what the reliability research actually shows.
Read the insight → InsightAI Adoption Is Behavior Change
Teams do not have an adoption problem because engineers resist. They have one because leadership rolls AI out like a corporate wellness program. The story of one risk-averse architect, and the single question that moved him.
Six things that actually work, and why a skeptical senior engineer is signal rather than obstruction.
Read the insight → InsightClarity, Not Proximity
I have led remote engineering teams since 2012, not 2020. Distance does not kill teams. Ambiguity does. Colocated teams survive vague ownership because someone overhears the confusion, and distributed teams never get that accident.
Why the fix was never more meetings, how RACI decides who attends, and what RANCI is.
Read the insight → InsightDriver, Not Doer
I passed on two senior candidates and told the recruiter why. Both could describe what they built from a spec. Neither had owned a problem end to end. As AI absorbs the implementation work, the executor profile is worth less every quarter.
What I screen for, plus Indeed Hiring Lab data on a senior-weighted market that is still well below its pre-pandemic level.
Read the insight →How These Connect
These are six arguments about one shift. Generating code got cheap, so the value moved to deciding what should exist, reviewing what came back, and owning the consequences. Everything downstream of that follows: review becomes the constrained resource rather than implementation, agents need direction instead of specification, adoption becomes a trust problem rather than a licensing one, and the engineer you want is the one who finds the next step instead of waiting for it.
If you want the same arguments with the architecture attached, the case studies walk through specific production systems and the trade-offs behind them. The about page covers the track record, and contact is the fastest way to start a conversation.
Looking for a senior engineering leader who can ship AI in production?
Open to a full-time senior engineering leadership role. No AI theater. Measurable outcomes only.
Hire Me →