AI writes the code now. Accepting it without really understanding it doesn't make anyone worse at coding today. It shows up later, the day a bug appears that the AI can't fix and the instinct to find it by hand is gone.

This is a real, measurable problem, not a hunch. Anthropic researchers Judy Hanwen Shen and Alex Tamkin's 2026 study, How AI assistance impacts the formation of coding skills, found developers who over-delegate score measurably worse on comprehension tests of code they'd just worked with minutes earlier. Lee et al.'s 2025 CHI paper, The Impact of Generative AI on Critical Thinking, found the same pattern across knowledge workers broadly: the more we trust AI output, the less we scrutinize it.

This talk presents three small, reusable habits to fight this directly, each open-sourced as an AI-assistant skill usable in Claude Code, Cursor, Windsurf, or Gemini CLI:

Align: staying in genuine sync with the AI while building, understanding the why before accepting anything, and turning unfamiliar concepts into real learning as they come up.
Assert: building the confidence to present and defend a decision in front of clients and senior engineers without flinching, proving understanding holds up under real scrutiny instead of just private review.
Reinforce: retaining what was learned over weeks through spaced, interleaved review, so it doesn't quietly fade.

Together, they keep a developer from quietly losing their edge, and instead moving toward someone who keeps growing alongside the tools they use.

Expect a live demo, the data behind the problem, and an honest account of what's proven and what's still an open hypothesis, because this problem is new enough that nobody has the full answer yet.

Attendees will leave with:
A concrete, research-backed explanation of AI-driven skill loss
Three reusable habits, not willpower, for staying sharp while delegating more
An honest view of what's proven vs. unproven, and how to test it on your own team