Gavin JuddIndex

CLI

Git Mog

Compares public GitHub profiles for entertainment: one profile gets a read, and two get a battle. Here's one battle as the terminal printed it.

GIT MOG
◆ @geohot WINS 65–58 · PUBLIC EDGE · LIMITED READ
Coverage: @karpathy 40% · @geohot 59%
» geohot ships versions. karpathy ships intentions. [1]
THE FIGHT
SHIPPING         10–65  @geohot · Project upkeep: 8/42–0/23 [2]
ACTIVITY         89–71  @karpathy · Last public update: 35d ago–0d ago [3]
MAINTAINABILITY  100–83  @karpathy · Largest source file: 32 KB–257 KB [4]
THE READ
@karpathy — Tests: 18 per 100 source files in inspected trees.
  34 larger projects have gone quiet for a year. [5]
@geohot — Tests: 16 per 100 source files in inspected trees.
  23 larger projects have gone quiet for a year. [6]
CODE QUALITY · PREVIEW
Not enough TypeScript/JavaScript for a useful sample.
Not used in the battle score.
RECEIPTS
[1] Versioned releases — 0 vs 2
[2] Project upkeep — 8/42 vs 0/23
[3] Last public update — 35d ago vs 0d ago
[4] Largest source file — 32 KB vs 257 KB
[5] @karpathy — Test-file density 18 per 100 source files; Project upkeep — 34
    larger projects quiet for a year
[6] @geohot — Test-file density 16 per 100 source files; Project upkeep — 23
    larger projects quiet for a year
More: --details · --receipts · --share x
Rematch: gitmog geohot karpathy
Next: gitmog geohot <handle>

An earlier experiment used local models, but they tended to pick whichever story plan came first, and almost none of their rewrites made the cut. The model layer came out, and the current version uses rules and prewritten lines.