AI Makes Sitting With Discomfort Feel Shameful
Sitting with discomfort has always been a big part of my learning journey. Mulling over problems, sketching on paper, doing Google searches, reading books, reading the documentation, listening to podcasts, refactoring code, writing blog posts, paying attention to the pain caused by incorrect abstractions, weighing pros and cons, and even continuing to use technologies and products that other people have deemed "dead" — it's all about sitting with my discomfort, paying attention to my feelings, and trying to evolve. And as much as that's always been a struggle, it was a struggle that felt rewarding. But now, in the age of AI (Artificial Intelligence), sitting with discomfort feels shameful.
Earlier this week, I wrote a blog post about using SELECT ... FOR UPDATE in ColdFusion in order to apply row-locking in a MySQL database. That blog post took me three pre-work mornings to write; and, it was based on five pre-work mornings of technical exploration.
Before AI, this week-long journey would have felt joyful.
Before AI, this week-long journey would have felt like I was building up a base of knowledge that would pay me back for years to come.
Before AI, this week-long journey would have felt like it was building character, perspective, and "taste" that would color everything that I did moving forward.
Before AI, sitting with the discomfort of learning about "read locks", trying to find the right patterns in my code, refactoring and then refactoring again — it would have felt like a virtuous battle of determination from which I would almost certainly emerge a better and more learned craftsman.
But AI has changed everything. Beneath every moment of beautiful agony, there's now a hissing background noise — a persistent voice that's continuously accusing:
"What the fuck are you doing?!"
"Why are you wasting your time?!"
"Why aren't you prompting an AI agent?!"
"You're becoming so fucking irrelevant!"
The technological zeitgeist tells me that I should be 100x'ing my productivity; that I should have 10 different side-projects in the works; that I should be running agents overnight; that multi-tasking is the only way to be successful in the new world order.
And it's made me feel ashamed that I'm taking time to sit and think about when a database read should be FOR UPDATE vs FOR SHARE. And I hate that this meditative practice that used to bring me much joy now brings me shame.
Even now, as I write this blog post, I'm feeling shame. After all, who even reads any more? Who even uses the web any more? Aren't we all just supposed to be publishing content in Markdown files so that the Agents can easily read it for us? I can't believe I'm sitting here trying to process my feelings when I could be prompting an agent. What a fucking waste of time!
In fact, why didn't I just prompt an agent to write this blog post for me? I could've been done in two minutes and with better AEO (answer engine optimization) designed to juice my site's visibility in ChatGPT. What a fucking waste of time!
Year ago, sociologist Sherry Turkle said, "If you don't teach your children how to be alone, they only know how to be lonely." I used to think this was a powerful statement about the importance of sitting with discomfort. But in a world where you can marry a digital avatar and scroll endlessly on TikTok, perhaps being alone is just as anachronistic as learning about database row locking.
I've vented long enough, time to get back to my digital needle-point or digital gardening or however we'll eventually start referring to the "hobby" of coding.
Reader Comments
This poetry you just wrote cannot be written by AI. It's such a beautiful and perfect description of reality.
Accurate on all ends. Feels like if we are not working on multiple projects, then we should be doing the Game of Thrones walk of attonement.
Thank you for sharing your emotions on this topic.
AI is a digital sledge hammer. It's not the right tool for finishing nails.
I feel you! And I loved that you used the term digital needle-point! Made me laugh 🤣
@Angel,
Thank you for the kind words. And as an aside, the
Shame!GIF from GoT is one of my favorite GIFs. Fun fact, the bell lady in that scene is the club owner (Hannah Waddingham) from Ted Lasso.@Chris,
One thing that I hate about this whole situation, is that I feel like so much of my angst is strictly the marketing / social media of it all. But the problem is, I have trouble believing that there's nuance to the things people say when they express no nuance at all.
Hey, if everyone's prompting agents all the time, then nobody will remember how to actually write code. So knowing how to write code and write words for that matter, will continue to be valuable.
I just read this in Feedly. I blog myself to express ideas and keep track of my own journey. I read your blog because it's interesting. I believe AI would turn it into something uninteresting. Your voice and point of view is what I care about. I don't even use ColdFusion. 😎
@Glen,
"blog myself to express ideas" ... this is where I'm at as well. Which is especially interesting in the age of AI because it really puts that to the test. When I write now, I have to be totally OK with the idea that its very possible - maybe even likely - that no one will ever read this. I've always told myself that this is the case; but when it's the reality, it has a different feel to it.
I just remind myself, this is for me, this is for me.
Dear Ben,
I am writing this correspondence so that you may know that I have read your blog entry. 😄
Also, no one wants to read AI slop blog entries, so thank you for continuing to write yours by hand!
Sincerely,
Pete
Ha ha, thanks Pete 🙌 will continue to keep this a slop-free zone.
How interesting, because I have the exact opposite experience!
It's whenever I obtain the answer outside of my myself, most especially when I get a fairly direct answer from an LLM that works, THAT is when I feel the most shame.
Not because an LLM gave me the answer necessarily, but rather because I feel like I missed out on an opportunity for growth and learning. Getting the answer without the friction means that I've missed an opportunity to cultivate real understanding, because friction is the only way knowledge can become actual, tangible, understanding.
I've realized I'm becoming some kind of Cognition Advocate. 😅 I have no problems using LLMs, but I'm highly selective and precise about when, where and how much I do. I still code without them a significant portion of any project and I never use them for writing (but I sure do love having them look for typos and grammar, while leaving my original text un-touched).
You might enjoy this article I just wrote, it has some tie-in to yours, but kind of from the other angle:
Agentic Coding is a Trap | Remaining vigilant about cognitive debt and atrophy.
Hey Ben,
I love this post.
AI will never replace craft. This applies to software development, graphic design, music, illustration, or anything creative that can be digitized.
Instead, AI replaces production at a high rate of speed.
The difference is subtle and hard to grasp.
All I can recommend is don't stop doing what you do.
AI will give you its 2 cents if you ask it. But there is no law that says you should or shouldn't.
If you do, you just need to clean the slop.
Hello Ben,
Been a long time since we took that Intro to Comp Sci class back in the day. I haven't commented until now, but I have been reading you occasionally. You are not alone in your thoughts, particularly amongst those of who've been in this industry for a while.
What's particularly striking to me is how cynical and nihilistic much of the rhetoric around AI has been from the very people pushing it, with their constant warnings of severe disruptions and threats to our livelihoods. I'm not sure whether they really believe what they say or see it as some sort of twisted marketing ploy, but either way, it is surely contributing to these conflicted feelings that many of us are experiencing. I found this post about the effect it is having on our mental health particularly insightful: https://www.linkedin.com/pulse/we-should-talk-how-ai-revolution-impacting-our-mental-erik-nygren-gc0pe/
Like you, I've enjoyed working in this industry partly because of the friction and discomfort. And I could be wrong, but I suspect that as the full ramifications of what AI brings to the table become clearer, there will still be a need for people like us who actually understand what's going on beneath the surface, particularly with the insane amount of code that is now being generated with less oversight. Perhaps we will end up akin to the Cobol coders who had to clean up Y2K.
@Alex Alexiou,
I'd go a step further and say this profession has historically required a temperament that tolerates both.
The ability to sit with a problem for hours, days, even weeks...scratching your head, backing up, trying different angles, failing repeatedly until finally...something clicks.
AI has flipped that script by dramatically reducing both friction and discomfort. With that, I think we'll see a new breed of developer emerge.
But even now, it feels like AI handles 80%, maybe 90% of the work remarkably well. That last 10% is where things get weird, stubborn, and often frustrating. You either continue to prompt your way out of it, or if you're us...you dig in.
And I suspect that's where those of us who've built a tolerance for discomfort still have an edge.
Wonderful and thoughtful post, Ben. I feel similar thoughts about the new AI environment in my own blogging experiences. It's a strange new world that we live in!
Take care,
Gregory
@Alex, @Chris,
A big part of the frustration in all the rhetoric is that, as you guys are saying, AI is actually really helping in a lot of places. I enjoy using AI to help me think through a problem or to help me more clearly articulate a friction I might be feeling in the code. And for really rote stuff, I do like that it can make bulk-changes.
But, the real problem is one of scale. There's so much messaging that if you're not using AI for everything then you're doing it wrong. Which negates all of the positive feelings I might be having about how I already use AI; because apparently it's never enough.
So here's an interesting context, which I think speaks to a lot of this. I have a side-project - Big Sexy Poems - that I've been using as a playground for new ideas. And, I want to bring a number of those ideas back into the coding of this blog.
Now, I could probably have AI help with that (hey, go look at this folder and help me apply those patterns to this folder). But, I actually want to do it manually. And for a specific reason. I want to feel the friction. Since I'm talking about bringing over patterns from one context to another, I feel like I want to do it manually so that I can see/feel/experience where the pattern doesn't quite fix. This will give me a hook to stop and think about why something doesn't fit, what assumptions I made that turned out to be wrong, and how I can evolve my thinking to make the pattern better or more flexible.
If I just ask AI to do that for me, then I miss out on all the learning and the pain that prompts me to keep evolving my understanding of how the application works and how the code fits together.
I strongly believe that we can only learn from our own mistakes (I've given up on trying to learn from other people's mistakes - I'm frankly not that smart). So if I remove too much friction, I'll almost certainly remove the learning.
@Lars,
I really like your article. I think you echo much of my own sentiment about the power of friction as the driving force for building taste and understanding. The notion of having to "think in code" also really hits hard for me. I can't tell you how many times I've thought about an idea as being good, even wrote up a doc for it, then the very first line of code I go do write, I'm like — wait, this is a terrible idea. There's something about writing code that just becomes critical to how you think about application architecture.
I was just thinking about recent studies / thinking about REM Sleep. As medicine evolves, we are finding that the brain is lot more "plastic" than we used to give it credit for. Not only can neurons regenerate, but areas of the brain can basically take over for other areas' responsibilities. It's kind of bonkers. And the speed with which unstimulated brain parts can devolve is somewhat shocking.
So one theory is that REM sleep is just a way prevent the visual cortex of the brain from being inactive for 8 straight hours of sleep (if you're lucky to sleep that much). Because even with such a short period of time, they can see evidence that the brain will detrain on its visual abilities.
Basically it's the old adage, "Use it or lose it". Only it happens a lot faster than we may have thought.
Was just watching a YouTube presentation by the Primegan and he used the phrase, "Toxic Productivity". I like that.
@Ben Nadel,
Love the theory related to the purpose of REM sleep. I had not heard that one before, but it makes sense.
We are the old guard of code-smithing. The new generation of programmers raised on the nipple of AI will not care about how the code reads, looks, or is assembled. They will solely evaluated on whether it works or not. And as much as it hurts me to say this...maybe they don't need to care. Maybe, this is progress. Maybe someday AI will just write byte code. Maybe.
Post A Comment — ❤️ I'd Love To Hear From You! ❤️
Post a Comment →