Research indicates that using LLMs causes the accumulation of "cognitive debt" - individuals take intellectual shortcuts. The LLM generates a wall of text, the operator might not even read it, yet a false impression is created that "the job is done." This affects research (not verifying sources - e.g., LLMs citing non-existent studies - ouch, politicians!), coding ("I don't read what I push to production"), and - most importantly for this session - learning.

What if LLMs are tools that simultaneously elevate us to new heights and lead to terrifying cognitive decline (and everything in between)? Saying LLMs are "just a tool" is all well and good, but how do we use them in practice to maximize those "heights" and minimize the decline?

As the cost of generating text and code plummets, the relative value of understanding and decision-making increases.

We will revisit the foundations of cognitive science and dust off the principles of "critical thinking" to apply them to "learning with LLMs." We will explore various techniques, methods, and tools for building personal understanding and mental models of specific topics. This includes verification methods, techniques for "drilling down" (exploring the layers of depth - and knowing where to stop), and many other practical tips.