How Much AI is Too Much AI?

AI has never been easier to use, but while everyone is enjoying tiny speed gains, we are quietly offloading our actual thinking. It's happening everywhere, especially in development when we blindly push code we didn’t write. This talk isn't a generic anti-AI rant; it’s a reality check on how over-dependence changes our brains. We are already seeing the practical "resistance" in open source, with organizations like Wikimedia and sktime actively restricting and even banning unverified AI pull request affecting FOSS policies.

What the session will cover:

1. The AI Slop Crisis & Codebase Decay: How mindless "vibecoding" to pad a green GitHub profiles degrades code quality, burns out the code reviewer, and creates severe technical debt.
2. The 17% Brain Tax (The Cognitive Debt): A study showed that AI users scored 17% lower on understanding code they just wrote. Skipping the "struggle step" creates a hidden "cognitive debt", code passes tests, but nobody understands when it fails in prod.
3. The 3 Levels of Drifting: The quiet slide from Level 1 (AI handles boring syntax), to Level 2 (reviewing AI logic instead of writing it), to Level 3 (complete replacement, shipping code that only god knows would work).
4. A Rigorous AI Usage Protocol : Practical habits to stop the drift. We will cover solid strategies like writing unit tests before prompting to lock down requirements, using semantic code search to anchor prompts in your existing architecture, and treating AI outputs as unverified legacy code.

How it could Benefit the Listeners:

1. A Smarter Dev Mindset: Walk away with an educated mindset on actually utilizing AI without being an AI addict. Using AI is absolutely necessary today, but how you use it is the key.
2. An Honest Reality Check: Get a clear mental map to introspect at their own workflows and catch themselves before they drift into dangerous over-dependence.

1. Awareness about cognitive offloading and its impact
2. A practical AI usage guideline for developers
3. Understanding of when to use and when not to of LLMs
4. A clear mental model