Wile E. Coyote at 40

In the early months of 2026, when I got really hot and heavy into Claude Coding, I had a new and unpleasant experience: a visceral sensation of AI fear. While I had been using various AI coding assistants back to 2022, and had of course been reading various takes on how AI was going to automate all the jobs for about as long, I had never really felt it. But in February 2026, that changed.

That wasn't my first impression, though; it started though as pure fun. I had taken over as head of Product at Precedent in October 2025, and the transition meant roughly a 2-month break from writing code. Then, over the holidays in December 2025, I started playing with Claude Code using the recently-released Opus 4.5 model. I began by prototyping some new features and getting shockingly good results; pasting the body of Asana tickets into the CLI generally resulted in fully working implementations, in minutes, on the first try. As I gained more confidence this went from rapid prototyping to just straight shipping features as quickly as I could define them. By mid-January I had burned through my backlog of (relatively simple, straightforward) feature tickets.

A month or so later, I sat down to do some cohort analysis on Precedent's customer base; I wanted to answer some questions about how volume, retention and turnaround time behaved vs other variables. In 2025, I had done similar analysis using Cursor and iPython Notebooks, and felt like a data analysis god. What would have taken days in 2015 took hours in 2025. And I didn't need to type out 100 opening and closing brackets to manipulate pandas DataFrames!

In Feb 2026, I just fed Claude the raw data and prompt explaining what I wanted, and got results a couple minutes later. A few quick checks showed everything lined up. What took hours in 2025 took minutes in 2026. But I didn't feel like a data analysis god. I felt like Wile E. Coyote, having run off the cliff, just now looking down.

This model/harness combo was great not just at implementing webapp features, but could produce really high-quality data analysis in one shot, including data cleaning, transformations, building and interpreting linear and logistic regression models. I had spent years mastering these skills, and had envisioned they would guarantee my job security pretty much indefinitely, regardless of what new frameworks or trends emerged in the tech world.

I kept thinking "This thing can do everything I'm good at."

In retrospect the difference is clear: when I used AI for something I wasn't good at (writing typescript) it was fun; when I used it for something I was really good at, it was scary.

I remember hearing about factories shutting down in 2008 and lots of talk about "retraining" and "reskilling." I didn't think too much about it at the time. I certainly wasn't rooting for anyone to lose their jobs but also didn't feel too much sympathy; those jobs seemed shitty anyway, so why wouldn't those middle-aged factory workers jump at the opportunity to do something more stimulating? Easy to say when you're 23 and learning new things every day anyway. Now, myself 40, witnessing the plummeting value of skills in which I've invested years of time, I get it. It's terrifying.

Fast-forward to April 2026. Precedent shut down, but I landed a new job pretty quickly. I still don't feel 100% great about Claude and Codex continuing to encroach on what I thought was my territory, but my outlook has improved a lot since February. One thing that has legitimately helped me is framing my own situation as a Pascal's Wager.

Will AI make my hard-earned technical skills totally obsolete? Maybe. While it's comforting that I was still able to land a new job quite recently, I don't discount the possibility that in a few years, there just won't be a need for humans to do data analysis, software engineering, or basically anything that I'm really good at doing on a computer. And if I knew for certain that were the case, I should probably pivot hard and learn how to plumb or something.

But what if AI is not going to make my skills obsolete? What would that mean? It means my skills are still valuable, though maybe they need some adaptation, some fine-tuning. Maybe I'm like a veteran athlete at the tail end of my career, and while I can no longer run fast enough to stay on the field, I still know the game better than all these fresh recruits. Maybe I could become a coach, maybe a GM, shit maybe I can get up in the broadcasting booth? Maybe I'll even be better in one of these roles than I ever was as a player.

If I assume AI is not going to completely replace me, what can I do to maximize my value? Well, I can:

• Seriously invest in understanding the newest tools
• Shift from purely technical work (writing the code) to adjacent work (architecture, defining requirements, indeed stakeholder management)
• Publish thought leadership blog posts?

I stand to potentially gain a lot by acting as if my skills will remain valuable, and if I turn out to be wrong, I haven't lost much. In the moderate scenario, where AI replaces basically all sitting-at-a-computer jobs but leaves room for plumbers and nurses and such, I just wasted a bit of time learning how to properly manage agent context. In the extreme scenario, where AI replaces all human labor or just takes over the world generally, we're all either fucked or saved anyway, regardless of what I do with my time for the next few months or years. Leaning into AI is therefore a "heads I win, tails who cares" bet, and those are usually good bets to take. The stakes are lower than in Blaise Pascal's version, but the game theory is the same.

I still get the occasional AI fears. But this framing honestly does make me more optimistic, and so far (admittedly only a few months in), the bet seems to be going in my favor.