The fourth and culminating essay in Stuart McClure's AI Exposed series confronts a truth that is inconvenient precisely because it cuts in both directions. AI sceptics are wrong to dismiss the technology as hollow hype — the progress is real and accelerating. But AI optimists are equally wrong to assume the trajectory is linear, the risks manageable, or the outcomes predetermined.
The inconvenient truth is that we are building increasingly capable systems that we do not fully understand, deploying them at speeds that outpace our ability to evaluate their failure modes, and doing so within economic incentive structures that discourage the kind of careful verification the moment demands.
Stuart draws on his experience as a pioneer of AI-driven security — Cylance became the fastest company to $100M ARR in cybersecurity history precisely because AI-native threat prevention genuinely worked better than legacy signature-based approaches — to make the case that genuine progress and genuine risk coexist. Neither triumphalism nor nihilism is an adequate response to that duality.
The essay serves as both a reckoning and a roadmap. It reviews the major claims that dominated AI discourse through the preceding years, scores them against what actually happened, and offers a framework for thinking more clearly about where AI delivers and where it disappoints. For leaders making consequential technology decisions, it is required reading.
If the earlier essays were diagnostic, this final one is prescriptive: here is what honest engagement with AI looks like, here is the discipline it requires, and here is why that discipline ultimately produces better outcomes than the alternatives.