in which the machine attempts to grade my emails
Roughly, "good" emails are rated by how useful the content is, with an "average" post being around +110. Separately, "bad" emails are rated by just how painful (political, distasteful, etc.) it is, with a logarithmic scale somewhat equivalent to the positive one (so -21dB for an email as bad as an average email is good).
Raw Score (0-300):
Base: +110 for "average" useful content
Technical/project updates: +30-50
Documentation/links: +20-40
Novel insights: +20-40
Clear structure: +10-30
Signal Rating (-50 to 0 dB):
Base: 0dB for neutral content
Personal drama: -5 to -15dB
Political content: -10 to -20dB
Meta-commentary: -2dB per instance
Coded references: -1dB per instance
Stream-of-consciousness: -3dB per topic shift
Color tag interpretations:
red: -5dB (drama/emotion/politics)
xantham: -3dB (meta-commentary)
yellow: -1dB (coded references)
green: +5 raw (technical notes)
blue: +10 raw (positive conclusions)
mogue: -7dB (personal/relationship)
hazel: -2dB (mild complaints)
teal: +15 raw (structured content)
orange: +10 raw (project plans)
gray: -4dB (oblique political refs)
violet: +20 raw (formal/technical)
💡 it doesn't like the "red", and does like the "orange". Nobody's perfect.