This is Part 5 of 5 of my 2025 AI Wrapped series. This post covers key lessons for myself and how I think with AI.

AI isn’t just a code-writer. It’s an intern analyst—one efficient at finding connections across massive datasets, surfacing relevant research, and compiling baseline information. But just like any intern, it requires verification. When AI suggests something, that’s the start of an investigation, not the conclusion.

“But using AI makes you dumber,” they say. Counterpoint: Widespread adoption of GPS means a whole generation of people can get to their destination faster without paper maps. Thinking with AI has surfaced real arXiv research papers and McKinsey publications that were invisible to me with traditional search and exposed me to entirely new concepts.

AI doesn’t make me smarter; it makes conversational exploration of topics and ideas possible. This conversational approach allows me to mind-map my existing knowledge onto new topics. So, no, I don’t feel that AI has changed my approach to curiosity or learning. What it has helped me do is connect my curiosity to primary sources.1

Whereas in the past I would use Wikipedia or traverse an ever-worsening search results page, now I can converse with AI and instantly get a reasonable baseline of information. Understanding how generative LLMs work is critical to understanding that AI doesn’t know anything. It generates the statistically most likely response based on the source material in its training data; even so, I’ve found this generated baseline information reliable and more correct than not this past year.

I acknowledge that every AI company has plundered humanity’s copyrighted knowledge, and that’s worth reviewing. This series is an observation that generative LLMs are here and act as prisms through which we view our world. They reveal patterns that weren’t previously observed, make connections not previously made.

AI doesn’t replace your thinking; it provides another lens through which to view the world. And just like any lens, sometimes you gain more clarity by inspecting what’s in front of you without it.

When I look back on 2025, I see that AI made programming viable for engineering managers again. But more than that, AI has evolved past the limitations that made skepticism reasonable—context windows have expanded, responses have become reliable, sources have become verifiable.

If you’re still citing studies that AI doesn’t boost productivity,2 consider this: computers replaced NASA’s human calculators, CAD replaced human drafters, Excel replaced human ledger scribes. But scientists, architects, and accountants are still highly trained, skilled professionals. The tools didn’t eliminate judgment; they shifted where professionals spend their cognitive energy.

AI doesn’t change what engineering is—it changes where engineers spend their time.

We’re still in the early days of this transformation. The knowledge workers who integrate AI into their work will compound their effectiveness the same way previous generations did by mastering new technologies. AI is the next tool in that progression.

Significant Revisions

  • Jan 21st, 2026 Originally published on https://www.jsrowe.com with uid 61BBBCC3-9C95-4378-A540-86C18FE9BC78
  • Dec 16th, 2025 Initial rough draft created.

Footnotes

  1. You don’t need domain experts to explain basic concepts. Starting with AI means that conversations with people have more depth when they do start. And besides, not everyone is available all the time. ↩

  2. GDP and productivity gains from previous tool transitions didn’t show up until companies fundamentally changed how they accomplished work. ↩