Developer reflects on how AI has erased the manual scaffolding bottleneck that once slowed throwaway prototypes. A year ago, setting up a new repo meant writing boilerplate before testing ideas; today AI agents generate project skeletons, wire up dependencies, and produce runnable code in minutes. The shift lets him move from “I wonder if…” to a working demo almost instantly.

Recent GitHub activity illustrates the new tempo: repos named Sakoa, a progressive systems language with an effect system and multiple backends; Kato, a hybrid notation format; Seal, a CLI that stores secrets in OS credential stores; and several UI components like Plim. Most of these prototypes compile, pass tests, and some are evolving toward full projects, something that would have taken weeks before.

Measuring his own workflow, he finds AI agents deliver roughly 4x faster time‑to‑PR, though occasional odd outputs offset gains. The speed boost expands the scope of work: refactors and experiments that once sat in “nice‑to‑have” now fit into an afternoon, and internal tooling improvements have cut codespace boot times by ~50%. He balances this by carving out manual coding sessions to keep his fundamentals sharp.