I want to be a Journey Programmer Again

At the same time, at least at the moment, this feels like just another tool. I'm old, started programming in the early 80s. Basic->Asm->C->C++ (perl-python-js-ts-go). Throughout my life things have gotten easier. Drawing an image on my Atari 800 or Apple II was way harder than it is on any PC today in JavaScript with the Canvas API or some library like three.js. Reading files, serialization, data strcutures, I used to have to write all that code by hand. I learned how to parse files, how to deal with endian issues, alignment issues, write portable code, etc but today I can play a video in 3 lines of JavaScript. I'm much happier just writing those 3 lines than writing video encoders/decoders by hand (did that in the 90s) and I'm much happier writing those 3 lines than integrating ffmpeg or some other video library into C++ or Rust or whatever. Similarly in 3D, I'm much happier using three.js or Unreal or Unity than writing yet another engine and 100+ tools.

ATM LLMs feel like just another step. If I'm making a game, I don't want the AI to design the game, but I do want the AI to deal with all the more tedious parts. The problem has been solved before, I don't need to solve it again. I just want to use the existing solution and get to the unique parts that make whatever I'm making special.

I guess he must have started programming a short time ago, if he can say that. LLM programming tools have just now been introduced.

English is just too poorly-specified. Programs need to be able to know exactly what they're supposed to do next, what their output is supposed to be, etc. Even humans need to ask each other for clarification and such all the time.

If you want to use English to specify a program, by the time you've adjusted it to be clear and specific enough to actually be able to do that...it turns out you've made a programming language.

Here's a typical scenario: you're a well-respected senior engineer at your company. Say you're an E8 at Meta. You spend your days in meetings, write great documentation, and read more papers than most, which helps you solve high-level architectural problems. You’ve built deep expertise in your domain and earned a strong reputation, both internally and in the industry.

But deep down, you know you’re rusty with tools. You haven’t written production code in years. You’re solid in math and machine learning theory from all the reading, but you’ve never actually built and shipped production ML models. You're fluent in linear algebra and what not, but you don't know shit about writing CUDA libraries, let alone optimizing them. When you check the job specs at companies like OpenAI, you see they’re using Rust. You might be able to write a doubly linked list in Rust, but let’s be honest—you’d struggle to write a basic web service in it.

So switching domains starts to feel daunting. To say the least, you'll lose your edge to influence. Even if you’re willing to take a pay cut, the hiring company might not even want you. Your experience may help a little, but not enough. You’d have to give up your comfortable zone of leading through influence and dive back into the mess of writing code, fixing elusive bugs, and building things from scratch—stuff you used to love.

But now? You’ve got a family of five. You get distracted more often. Leadership fits your life better—you can rely more on experience, communication, intuition. Still, a part of you misses being a journeyman.

So how does someone actually make that move? Do you just bite the bullet and try? Stick to adjacent areas to play it safe? Join a company doing the kind of work you want, but stay in your current domain at first—say, a backend engineer goes to OpenAI but still works on infra? Or is there another path?