Vibe coding promises 10x productivity gains. PyPI doesn't confirm them yet

Enthusiasts of vibe coding and agentic tools say they have become 2x, 10x, even 100x more productive. One developer built a browser from scratch over a weekend. Skeptics ask a fair question: if that's true, where is all the new software?

The vibe coding phenomenon

Since 2023, AI assistants in software development have gone from a lab experiment to an everyday tool. Cursor, GitHub Copilot Workspace, Claude Code, Windsurf: new products ship every quarter. The principle is the same across them: the developer states an intent, a language model generates code, and the developer reviews and corrects it.

The productivity claims sound serious. They come not from beginners but from experienced engineers with years on the job: 5–10x speedups on routine tasks, an MVP in a week instead of a month, a solo SaaS product in a month instead of a quarter for a team. On social media this is no longer the exception but a genre.

The economic logic suggests that if development has become cheaper, there should be more of it. This reasoning assumes the world wants more software, so if it's cheaper to make, more of it will get made. If you accept that, a measurable "AI effect" should exist.

The PyPI test

PyPI, the central repository of Python packages, is an ideal yardstick for this hypothesis. It is large and stable, and its data is public and has been collected for years. Python is one of the main languages of the AI ecosystem, so if the effect shows up publicly, it should show up here.

Signs we would expect to see after 2023:

• growth in the number of new packages published per month
• more first-time package authors
• new thematic clusters (agents, LLM tooling)
• faster release cycles for existing libraries
• shorter time from first commit to publication

PyPI data is open and available through BigQuery. If the growth curve of publications turned sharply upward after 2023, that would be a weighty argument for the vibe coding narrative. If not, the effect exists but shows up differently than expected.

Why the data may stay silent

Even if the gain is real, PyPI may not show it. Most vibe coding projects never reach public registries: personal utilities, internal company tools, and client prototypes all stay private.

AI speeds up the start above all: sketching the architecture, generating the scaffold, writing tests. The slow parts, such as final debugging, documentation, and post-release support, stay just as slow. The result: projects start faster, but they don't finish more often.

Ambitions also grow. A developer used to spend a week on a simple script. Now, with the same week, they take on a full product with a UI, an API, and a database. The total volume of published code hasn't doubled, but the complexity of each project has grown. PyPI doesn't capture that.

"If we don't see the number of packages double, the speed gain is

being absorbed by something else" is roughly the central thesis of the skeptics.

What this means

Vibe coding is a real shift in how development feels. But public metrics don't yet confirm an explosive growth in the volume of software being created. Possible explanations: most AI projects stay private; the gains are redirected to more complex tasks; or we are at the start of the curve and the data will show up in a few years.

Either way, before taking the 10x and 100x figures at face value, look not at tweets but at repositories.