A 14-year software dev posted Tuesday that he let four interns vibe-code from day one — with rules. Two months in, the post hit 591 upvotes and 75 comments on r/vibecoding. The thread reads as a workshop manual, not a victory lap. Interns pair-program with the LLM; senior eyes review every PR; a written rulebook bounds what vibe-coding is allowed to touch and what it isn't. The comments are practical — tool choices, pitfalls, onboarding time. Nobody is arguing about whether vibecoding counts.

The tools stopped asking for permission

On GitHub, Onlook — the "Cursor for designers" project — crossed 25,000 stars this week. Easy-vibe sits at 6k. Neither repo is pitching a manifesto. Both ship a product: a canvas where a non-coder can build a real app by describing changes in natural language, with the code materializing underneath. The README reads like a product page, not a research claim.

On X, the genre has calcified. Builders post onchain tip jars deployed in fifteen minutes. Multiplayer browser games shipped from a single prompt. Internal tools written during a long flight. The posts don't argue that vibecoding works. They assume it and measure the output.

What changed

The 2025 discourse was about legitimacy. "Is this real engineering?" "Will the code hold?" "What happens when it breaks?" Those questions are still valid, but they've been answered in the only way that matters in this industry: somebody shipped it, somebody else copied the pattern, the pattern worked, and now it's a market. The interns post is evidence that the debate moved from the editorial pages to the training docs.

The rules part matters more than the vibe part. Every durable workflow in the threads has the same shape: scoped problem, explicit constraints, human review, shipped output. Vibecoding without rules is a demo. Vibecoding with rules is a junior engineer's onboarding path.

What to watch

Two signals for the next thirty days. First, hiring posts — when "vibecoding literacy" shows up in a job description from a company that isn't an AI lab, the framing has won. Second, the tool consolidation. Onlook, easy-vibe, Cursor, v0 and their imitators are all converging on the same product shape. Whoever owns the "builder who can't code" wedge in six months will own the next wave of indie launches.