Go experts: 'I don't want to maintain AI-generated code' (thenewstack.io)

Go veterans Dominic St‑Pierre and John Arundel used a recent podcast to unpack how AI code generation is reshaping the Go ecosystem — from a measurable uptick in Go usage (Stack Overflow: 16.4% in 2025 vs 13.5% prior year; TIOBE trends suggesting Go may overtake JavaScript) to the Go team’s explicit effort to make “Go better for AI” with integrations like LangChainGo and Genkit. They note why Go is especially exposed: its simple, syntactic surface and portability make it easy for LLMs to emit runnable binaries and CLIs, accelerating hobbyist and production experimentation. Their real‑world verdict is cautionary. St‑Pierre reports many low‑quality, AI‑generated repos and a collapsing acceptance bar that creates a deluge of mediocre PRs; his Claude tests produced code he wouldn’t want to maintain. That shifts engineers’ work from authoring to reviewing, refactoring and architectural control, raises onboarding/training concerns for juniors taught by AI outputs, and could increase demand for consultants—while also flooding the market with short‑lived projects. Arundel strikes a balanced note: more people will try programming, and bad AI code isn’t categorically worse than bad human code, but the net effect is a transitional landscape where tooling, review practices, and education must evolve to keep systems maintainable.