Don’t Listen to Lee, Listen to Chris Lattner
Who is Chris Lattner? The guy who developed LLVM (the compiler infrastructure), Clang (the C/C++/Obj-C compiler frontend), and the Swift programming language. Someone far more able than I to evaluate the achievement Anthropic made with their “mostly AI” coded Rust based C-compiler.
https://www.modular.com/blog/the-claude-c-compiler-what-it-reveals-about-the-future-of-software
You will hear themes I’ve been saying a long time. It is not “AI can code compilers by itself” but that there will be human+AI co-development.
"ThePrimeagen" gave his hot take - Anthropic lied - and pointed out all the ways the compiler sucked or wasn’t original or needed the existence of the gcc to crib from. And he’s right in terms of what he paid attention to.
What folks are missing is that Anthropic did not hire compiler development experts, team them up with the best agentic orchestration professionals in the world, and embarked on creating a breakthrough in compilers. They just tested their own model’s ability to work nearly autonomously for a couple weeks straight and develop a compiler, in Rust, that could compile Linux. And they succeeded.
Criticizing the compiler misses the point. Worse, it leads to missing the opportunity. But Chris isn’t going to do either.
Chris: “The most effective engineers will not compete with AI at producing code, but will learn to collaborate with it, by using AI to explore ideas faster, iterate more broadly, and focus human effort on direction and design”
I should sue him for plagiarism, right? Kidding, but I understand why “weird AI loving, kool-aid drinking, Apple fan” isn’t credible to some here. But Chris is a guy who knows what it is to build a compiler and he’s able to see the signal. It’s not “AI built a crappy compiler” - it was “imagine if my team of compiler experts used these tools”.
Chris: “Developments like the Claude C Compiler have changed how I think about engineering work and what I now ask from my team. Fully benefiting from AI tools requires a deliberate leap: habits formed over decades don’t change automatically, and organizations rarely transform just because better tools exist.
At the same time, we need to be pragmatic. AI systems are powerful but far from perfect. Progress comes from collaboration with AI, not abdication to it. The goal is not to remove humans from the loop, but to move humans into higher-leverage positions inside it.”
I write from the position of a consulting professional who makes a living selling my tech skills to clients. I saw GenAI from that lens and began retooling for what I was sure would impact my ability to stay relevant. Chris is a CEO looking at keeping his company viable into the future. Different guys, different perspectives - both reading the same signal and making plans.
Are you making plans?
ps: in addition to being fond of punch flavor Kool-aid — I’ve been working professionally in all things data related since 1983. I’ve navigated tech change after tech change for 42 years. I’m a principal level cloud data architect. I’m the “one throat to choke” on 7 figure engagements. I’m no Chris Lattner, but I’m not some vibe coder with a $20/mo ChatGPT pontificating on what I don’t know.
