Confirmed 90% prediction — "I don't write any code anymore." (Mar 2026)
Boris Cherny
Ships 10–30 PRs/day, zero manual code. Created Claude Code. (Feb 2026)
Andrej Karpathy
Coined "vibe coding," now full agentic — hasn't typed code since Dec 2025.
Jensen Huang
"The programming language is human. Everybody knows human." (Jun 2025)
The question is no longer whether AI will write code. It's what humans will do when it does.
What the Builders Say
Independent voices — no AI products to sell
LT
"The python visualizer has been basically written by vibe-coding. Is this much better than I could do by hand? Sure is."
Linus Torvalds shifted
Creator of Linux & Git · Jan 2026
DH
"You can't let the slop and cringe deny you the wonder of AI. This is the most exciting thing we've made computers do since we connected them to the internet."
DHH shifted
Creator of Ruby on Rails · Jan 2026
KB
"90% of my skills dropped to $0. The leverage for the remaining 10% went up 1000x. I need to recalibrate."
Kent Beck evolved
Creator of TDD & Extreme Programming · 2023–2026
DS
"AI slop is a Denial-of-Service attack on open source maintainers." Shut down curl's bug bounty over AI-generated fake reports — but acknowledges AI finds deep bugs no tool ever could.
Daniel Stenberg nuanced
Creator of curl · FOSDEM Feb 2026
JC
"'Coding' was never the source of value. Problem solving is the core skill. AI training on open source code magnifies the value of the gift."
John Carmack
Creator of Doom & Quake · Mar 2026
SW
"Our job is not to type code into a computer. Our job is to deliver systems that solve problems." Coined "Deep Blue" for the existential dread developers feel.
Training data bias favours popular languages (Python, PHP). But the consistent ordering across all 12 models reveals systematic structural challenges too — it's not just familiarity.
TS < JS
TypeScript scores lower than JavaScript despite having types. Types alone don't help if they add complexity.
C, Rust: worst
Systems languages score lowest across every model. Headers, manual memory, macros — implicit complexity that no amount of training data overcomes.
Qiu et al. 2025, LoCoBench, Salesforce AI Research — arXiv:2509.09614
Explicitness Without Complexity
Complexity kills AI
C, Rust score worst across every model. Headers, manual memory, macros — implicit complexity that no amount of training data overcomes.
The sweet spot
Simple structure + explicit semantics. Not types for safety's sake, but clarity for reasoning's sake. Explicitness that reduces complexity.
Training data bias is real — Python wins partly on volume. But the structural signal is consistent across all models: implicit complexity is the enemy. The answer isn't more types, it's more clarity.
If AI codes best in AI-friendly languages… maybe AI should help design that language. That's AILANG.