25 Years of Built-Up Work
I was talking to someone on HN recently. He told me he has more than 25 years of commercial development experience. Over the years, he’s been building his own library of solutions. It started with ActionScript 3 (yes, the tech that powered Flash back in the day), then he ported it to TypeScript. Along the way — a whole scatter of backend technologies. Bottom line: he has massive experience, and thanks to those old building blocks he can ship projects faster, cheaper, and at a higher quality.
As he puts it himself, quote:
“It’s very easy for me to launch new applications, services, or websites — ones that actually work and don’t have unknown vulnerabilities or garbage code.”
A few weeks ago, he decided to run an experiment with Claude. He asked it to write one of the components for his SPA framework from scratch. He spent 10 hours and barely managed to assemble 500 working lines of code. And most of that time was spent fighting over things like, “this component is missing a destructor.” He hated the code. Completely. And after enough complaints, the machine would respond with something like:
“You’re absolutely right, let’s do it the smart way!”
He didn’t even get to backend interaction.
Our conversation started when I said I was worried that in a few years even large corporations would start questioning the value of their own solutions. After Anthropic wrote a compiler that managed to build the Linux kernel, maybe their competitive moats aren’t that protected anymore. I suggested that at some point we might be able to assemble software at Photoshop-level complexity using AI tools. Say, modernizing a call-center system or a corporate CRM. Done by a single person who actually understands the problem — at a cost lower than a corporate license.
He had to disagree:
“Every piece of code I’ve written, I understand and can trust. That’s what allows me to build more and better things on top of it next time. What you’re describing sounds like resignation, laziness, solipsism, a desire to watch the world burn — or something along those lines.”
At work, I meet a lot of people with similar backgrounds. But the number of people who actually manage to use these new tools keeps growing. I personally spent many hours figuring out how to make them produce working code. And more importantly — I stopped treating transformer-generated code as dumb trash I need to shovel through. I started treating it as something important and new.
So much so that I began looking differently at my own old code — code I have a lot of sentimental attachment to. I admitted to myself that my code can become morally outdated. It’s still very, very valuable. Because the average code AI writes from scratch is disgusting. It really is statistically average garbage stitched together from billions of lines written by mediocre people. But my own code also becomes outdated — in terms of approaches and tooling.
Things that used to be impossible are now doable in much less time. I have more room for creativity and experimentation.
I don’t want to devalue the feelings of people who’ve built large, deeply personal libraries of work. Back in the day, carpenters probably treated their tools the same way. They hung them carefully around the workshop, engraved polished handles earned through hard labor, passed tools down to sons and grandsons. They were heartbroken if a son grew up clumsy and couldn’t carry on the family craft.
But progress kept going. Today you can cut a board with a soulless power tool — and we’ve lost that culture of craftsmen. At the same time, the number of cut boards in the world has increased dramatically.
AI changes how we relate to code. This may sound painful, but old work has less value now. On the other hand, the core goal of developers hasn’t changed. We still solve problems. We don’t write code for the sake of writing code.
If you’re interested in how AI can be integrated into your company, I’m happy to talk — just send me a message.
