Corrections!!
People have pointed out that the word "paradox" does indeed mean "counter-intuitive thing" rather than the a logical paradox like "this sentence is false." I have been laboring under the illusion that "logical paradox" and "paradox' are the same things! This is a disaster. Point for Jevons!

As Hank becomes more efficient, we will use more of him. All as Jevon has commanded

Guttenberg never imagined printing presses being used for instruction manuals for water heaters.

Woah Hank I appreciate the fact that you keep making these intros to the Connections videos longer and longer!

"I might be wrong about this" – Hank Green, like dozens of times in this video XD

“There are only four things. Possibly six.” - Hank Green

I'm a software engineer. I specialize in developer efficiency, workflow automation, code quality, tools, reliability, etc. Force multiplier stuff. So this feels very directly relevant to me.

I think it's important to keep in mind that there's the cost to write code, but separately there is the cost to maintain code. Code is tech debt the moment it's written. It rots and breaks, kind of like all the aging bridges all over the USA.

The world of software is changing a lot with AI, but I believe AI is still just one more tool, and we've spent decades using code to create tools to write code, and they help, but there are still serious limitations. AI lets us generate code, but that generated code tends to be a complex buggy mess, no matter how much people claim otherwise. And complexity itself is tech debt that slows everything down.

AI makes prototyping much, much faster. But actually building reliable systems doesn't feel that much faster than Pre-AI, so far. It's more like having advanced Google, where I can work more broadly outside my SME areas and still be somewhat effective.

I've been experimenting a lot with using AI. It is powerful for a lot more than just directly writing code. I'm using it to help me create new static analysis tools, to do AI code reviews, to do security scans of existing code repos, to help me test my own software, etc. AI itself is not really automation, but it's helping me build more automation. It does still take a ton of close supervision. It's not hands-off. The more carefully and closely I guide it, and the more work I put into context management (like through setting up code indexing tools for my agents to use), the better it does.

At the same time, AI is extremely dangerous because it's smart and stupid at the same time. It knows all the most advanced techniques but has extremely poor judgement about when NOT to do something. So it regularly creates disastrous situations with hugely negative impacts.

I'd also like to note, historically, measuring developer productivity has been a huge challenge. Measurements of lines of code written are notoriously misleading or perhaps entirely meaningless. There are stories (perhaps urban legends) of the best engineers having negative lines of code written, because they delete more code than they write. What does that mean in a world of AI? I'm not sure.

Disclaimer: these are just my own opinions and I'm not writing on behalf of my employer or anyone else. Also I'm a hands-on engineer trying to deliver software, not an academic or researcher.

"Unfortunately, you need to know" is like my whole life in 5 words

I'm reminded of efficiency of code when computers had limited memory.

When memory increased, code got less efficient and programs exploded in size

Near the end you asked, “What am I doing with my life?”

For what it’s worth, you’re helping at least one person rediscover hope, curiosity, and delight. I’d gotten to a place where those things felt pretty far away, and your videos have been gently but persistently reminding me that the world is still fascinating, that no effort is wasted, and that where I place my attention shapes how I feel, what I care about, and how I experience the world.

So… that’s one thing you’re doing with your life.

Thank you!

3:38 "The Jevens Counterintuitive Fact About Economics" doesnt have the same ring to it

Okay Hank Green I will learn what the Jevons Paradox is

2:37 there’s also the fact that any good capitalist would see “same profit, less costs” and say “no, more profit, same costs”

This reminds me of the quote from Alexander Graham Bell regarding his new invention.  

"One day, every major city in America will have a telephone"

I’m a transportation engineer and we deal with this all the time. We call it “induced demand”, and in a nutshell: almost always expanding the capacity of a road doesn’t decrease traffic congestion because you now made that road more desirable to drive on, and therefore more people chose to drive on it.

Hank: you can tell whether or not the code does the thing you ask it to do.
Me, a computer scientist: wait... no that's actually really hard.

"There is an upper limit to the amount of hot water people can use."

Teen girls (including myself at that age): "And I took that as a challenge."

My husband is a video engineer, he uses AI to code everyday. He was optimistic at first that using it would help take some work off his plate, but he works just as hard now if not harder because people on his team are demanding more features, and all the AI is doing is making more bugs more quickly. It’s like that comic “I’m doing 1000 calculations per second and they’re all wrong”

Infosec specialist here... The attack surface will always expand faster than we can figure out how to defend it. Our oldest maxim is that any system that can be designed by a person can be defeated by a person. Only now it's evolved to anything that can be designed by one system can be defeated by another system.

By far the best piece of speculation I've seen is Tom Scotts "2030: Privacy is Dead" talk that he did in 2015. Considering it was written before LLMs, it's felt almost scarily accurate so far.