A developer essay published on Yagnipedia is making a case that cuts against conventional wisdom: Go's strictness is a feature, not a liability, precisely because AI now writes most of the code.

The piece, posted under the handle 'riclib' and titled "Vibe in Go," starts from a simple premise. When AI systems generate 90% of a codebase, the question isn't which language lets humans write code fastest — it's which language catches machine mistakes most reliably. By that measure, the author argues, Go wins.

The argument centers on what riclib calls a five-layer defence: the compiler, the type system, explicit error handling, Go's enforced simplicity, and finally the human reviewer. The layers are ordered by cost. The compiler is free and instant, catching bad AI decisions before they run. Humans are expensive and slow, catching only what everything else missed. JavaScript, the piece argues, skips straight to that final layer. Every mistake the AI makes becomes the developer's problem to find.

The author illustrates this with a concrete failure mode: an AI generates a spurious wrapper type. In Go, it breaks at compile line 47. A human corrects it in five words. In JavaScript, the same type ships silently, spreads through the codebase over months, and eventually demands adapter layers to undo. One is a five-second fix. The other is a six-month liability.

To frame the stakes more precisely, riclib reaches for combinatorics. Twelve decisions per ticket, each with two to four alternatives, yields 559,872 possible decision paths. Most of those paths lead somewhere wrong. Go's compiler and type system collapse that space before any human reviews a line — not through intelligence, but through mechanical rejection of anything that doesn't compile.

The piece introduces the term "Vibe Engineering" to distinguish this mode of working from the more passive "vibe coding" — a phrase the article claims predates Andrej Karpathy's widely-cited February 2025 coinage by roughly six months, tracing it to an August 2024 Twitter thread. Where vibe coding implies accepting AI output, Vibe Engineering implies steering it. The human's job is to make five-word corrections to what four automated layers already filtered.

The implications for teams running agentic workflows are direct. As AI agents generate, review, and iterate on code with fewer human checkpoints, the language they write in starts to function as a quality gate in its own right. Go's compile-time constraints don't depend on the agent's judgment — they're a hard requirement the agent has to satisfy regardless of how confident it sounds. That property may matter more than ecosystem size or expressiveness as agent autonomy in software development keeps expanding.