AI has become impossible to ignore in software development. It promises speed, leverage, and productivity. It also produces shallow designs, fragmented models, false confidence, and a lot of very convincing nonsense. For teams that care about long-term system health, shared understanding, and conceptual integrity, it is reasonable to ask whether AI is doing more harm than good. Skepticism is warranted, and caution is necessary.

But rather than treating AI as a miracle or a menace, this talk shares hard-won lessons from hands-on experience using AI in real design and development work. The core claim is simple: the real constraint has always been discovery, sense-making, and alignment. Used carelessly, AI amplifies the wrong things. Used deliberately, it can meaningfully support those deeper activities.

The session explores where AI helps and where it actively gets in the way. Topics include working with non-determinism instead of fighting it, designing workflows that assume failure rather than correctness, using AI to explore design alternatives without fragmenting understanding, and recognizing which problem spaces benefit from AI assistance and which do not. Along the way, the talk surfaces evergreen principles that still matter even as the tools change, including clarity of intent, semantic anchoring, and responsibility for decisions.

This is not a talk about tools, prompts, or hype. It is about judgment. Attendees will leave with a clearer mental model for when AI adds leverage, when it creates risk, and how to integrate it into serious design work without surrendering thinking, accountability, or coherence.