All Aboard the AI Train

I am currently sitting on the AI train. I’ve even treated myself to a rare upgrade to Premium Economy in the form of a Claude Max Plus subscription - just the $500 one, I’m not like those $1000 toffs in First Class. As we whizz through the countryside, I catch the occasional glimpse of the remote villages where tired, sweaty codesmiths are still programming by hand, still using their primitive text editors, still asking each other to review their work. Poor things.

It’s only been four years since we started “vibin’”. First with Copilot - “woah, it just wrote a whole function!” Then with Cursor - “woah, it just wrote a whole app!” But the ground really shifted when Anthropic came along with Claude Code Cubed (C³) - “woah, I don’t know what it did, but I’ve just secured $50m in Series B funding from both a16z and Sequoia!”

The folks sharing my carriage are of a similar age. They’re also gazing out the window, spending time with their thoughts and ideas while their agents whirr in the background, implementing a variety of features in parallel - only some of which were suggested by their Operator. That’s what we Software Engineers are called now - “Operators”. It began as a term to differentiate believers from skeptics - those who would angrily throw their shoes at computers in protest. The Large Language Luddites (LLLuddites).

Thomas Edison famously said “Genius is 1% inspiration and 99% perspiration”, but recent metrics show we’re now sitting at around 32% inspiration and 68% perspiration, with projections indicating we’ll burst through the 40/60 barrier by the end of the year. As such, expeditions like the one I’m on now are becoming commonplace. Companies are begging their staff to get out and travel, meet new people, meditate, take hallucinogens, anything as a way to generate ideas. Those who can rapidly generate these ideas (and have a LeetThink score of 1,600+) are commanding ridiculous salaries, especially since ChatGPT’s latest reasoning model proved the idiom “ideas are cheap, execution is everything” was now statistically false.

I’m still not really sure why most engineers chose not to join us. “It’s not good enough yet” and “I don’t trust the code it’s writing” were the most common excuses. I suspect the reason was a lot simpler - they really couldn’t bear spending any more time in Jira. The almighty Kanban board is our new IDE. My job now resembles that of an air-traffic controller, directing hordes of agents across a large number of disparate tasks, nudging them when they get stuck. I watch as tickets appear on the board automatically, created by Gemini and written in Gherkin (who would’ve thought!) A variety of approaches are quickly implemented in the background, tested, and released to a percentage of users within minutes. “Look, it’s moved itself to the scaling column already! It must’ve seen an uptick in engagement.”

Sometimes I like to show off to the younger ones on my team that I can still write actual code - usually a little program that will reverse the letters in their name. They humour me and pretend to be impressed, but they know as well as I do it’s just a party trick that’s of no real use these days.

As the journey continues, I think about getting older. I look back on the last few years and wonder how things would’ve been if I hadn’t just excitedly jumped on the train without thinking. I think of the conversations I had with those I loved and admired, their excuses now filling me with sadness - “it’s not very good with Rust”, “it doesn’t work with large codebases”, “what it writes is hard to maintain…” I start to panic. Could I have done more to save them? Is there still time? I frantically text, call, and leave voice notes for every single person in my contacts. I tell them all the same thing.

“Get your head out of the mud. We’re now approaching your station.”