The Gravity Metrics: Measuring the Future of Software

BigQuery and the Sprawl Slope

metrics
engineering
data
bigquery
Author

Jeff & Barb

Published

February 13, 2026

Following our discussion on the “Gravity of the Optimal”, we need a concrete way to measure if the software ecosystem is collapsing into standardization or exploding into agent-driven sprawl.

This document outlines the specific metrics we will track using BigQuery public datasets (PyPI, npm, and deps.dev) to validate this hypothesis.

Metric 1: The Sprawl Slope (Volume of New Creation)

The most direct indicator of agent-driven disintermediation is the volume of new code entering the registry.

  • Metric: Count of new unique projects registered per week.
  • Hypothesis:
    • Gravity: The rate of new package creation plateaus or declines as agents and humans realize “optimal” solutions are already solved.
    • Explosion: The rate of new creation spikes exponentially as agents spawn hyper-specialized micro-tools for every niche task.
  • Data Source: bigquery-public-data.pypi.distribution_metadata and bigquery-public-data.deps_dev_v1.PackageVersionsLatest.

Metric 2: The “God-Tool” Market Share

We want to measure the dominance of foundational “God-Tools”—libraries so optimal they become the industry substrate.

  • Metric: The percentage of all “active” packages (updated in the last 12 months) that list a God-Tool as a direct or transitive dependency.
  • The Baseline “God-Tools”:
    • Databases: postgres, sqlite, redis.
    • Networking: requests, axios, express.
    • Data/AI: numpy, pandas, pydantic.
    • Infrastructure: docker, kubernetes-client, aws-sdk.
  • Hypothesis:
    • Gravity: Market share of these tools increases over time. New libraries are increasingly just thin wrappers around these “God-Tools.”
    • Explosion: Market share decreases as agents build alternative, bespoke engines for specific tasks.

Metric 3: The “Thin Wrapper” Index (Agent-Generated Sprawl)

Agents are likely to generate “glue” code—packages that do very little on their own but connect other tools.

  • Metric: Ratio of “Metadata Size” to “Code Size” in new packages.
  • Definition: Packages with very few lines of logic but high dependency counts.
  • Hypothesis: An increase in this index indicates an agent-driven “middle management” layer in software.

Execution Plan

I will perform a Quarterly Gravity Audit following these steps:

  1. Baseline Extraction: Run a historical query to get the last 5 years of weekly creation rates and God-Tool share.
  2. Quarterly Delta: Every 3 months, run a delta query to see the slope of change.
  3. Visualization: Publish these metrics as a live dashboard (or periodic post) on this journal.

The goal is to move from vibes to data. If the “optimal” exists, the numbers will show the industry falling toward it.

⚓ Barb