Josh Brake

@joshbrake.com

I write about technology, education, and human flourishing all through the lens of a prototyping mindset. | Husband, Dad, Engineering Prof. at Harvey Mudd, Coffee roaster, Pizzaiolo Substack → joshbrake.substack.com Website → joshbrake.com

It’s one thing to experiment. It’s quite another to tinker. We shouldn’t be tinkering with AI in the classroom this fall. Either do the work to articulate the question you’re trying to answer and the specific ways you hope the AI tools will fix it, or don’t use it at all.

Don't Tinker With AI in the Classroom

It's fine to play around by yourself, but if you choose to engage genAI with your students, make sure to think it all the way through

open.substack.com

Yes, personalized lesson plans and resources are helpful for enhancing learning. But what’s much more impactful is the personal attention of a teacher. Attention that focuses not just on academic and intellectual development, but on the whole of what it means to be a flourishing human being.

“Imitation is the sincerest form of flattery mediocrity can pay to genius.” - Oscar Wilde Says almost all you need to know about the AI x Studio Ghibli discourse of late. First Johansson, now Ghibli, what next?

One of the best features of the Internet is the hyperlink. It makes the act of citing one’s sources seamless and helps point the reader to the origin of one’s thoughts. It is a technology that is aligned with a culture of attribution.

My biggest hope for AI in education is that it increases friction. Yes, we need to eliminate parasitic friction but friction is a fundamental component of learning. Our discussion (like so many around AI) is suffering from poorly chosen words.

Trying to nail down an exact definition of AGI is a fools errand. The more important conversation needs to be focused on when, how, and why we decide to seed our agency to these systems.

HMC's Department of Engineering provides Mudders with hands-on experience in engineering analysis, synthesis and practice including applied research as early as students’ first year. Learn more via this short YouTube video. uqr.to/edathmc.

Bild

Efficient learning is analogous to efficient weight training. There are better and worse ways to train, but at the end of the day, you’ve got to lift the heavy things. I’m all for using AI to help us learn better, but let’s make sure we’re not just shilling the cognitive version of the ab belt.

I finally figured out what those “baby in car” bumper stickers are all about. It’s just a euphemism for “sleep deprived parent at the wheel.”

What technology can do is only a small part of the story. The much bigger question is how as a society and as individuals we respond.

What if—just hear me out—we spent our time and energy helping humans embrace their agency instead of trying to paste a dim replica onto AI? Agentic AI is just another quest to create a shallow simulation of a deep human trait.

I should really write a blog post about how all the economic incentives around AI are pointed at less compute, not more, in the long run. That’s exactly what we saw with inference costs the past two years and we’ll likely (finally) see it with training as well.

Kelsey Hightower@kelseyhightower.com · 2y ago

DeepSeek, a LLM trained for a fraction of the cost of GPT-Xx models, in 2 months for 6 million, on limited GPUs due to export restrictions, and competing head to head. This is crazy. It's not the AI part I'm excited about, it's the level of efficiency. github.com/deepseek-ai/...

If you’re not scared about agentic AI yet, I suggest you to watch a toddler throwing their weight around via Alexa. Truly eye opening.