The Mind Reader
Preamble
The Mind Reader is deliberately literal. It reads handwriting, extracts commitments, and turns the morning’s fog into a score I can argue with less easily.
The Translation
The first pass is automatic. I drop a photo or scan into _RAW/morning-pages/. A watcher sees the file, runs Apple Vision OCR, checks confidence and word count, then writes the page into the private Obsidian vault.
Bad scans fail. Good scans become markdown. That quality gate matters because a false reading of a private page can become a false account of the day.
The second pass is deliberate. I run /score-morning-page inside Claude Code. The model reads the latest page and extracts commitments. Each commitment gets three scores:
| Variable | Range | Meaning |
|---|---|---|
W | 0.1-2.0 | weight: how much the commitment matters |
F | 1.0-2.5 | friction: how much resistance it carries |
P | 0.5-1.5 | precision: how clearly the action is named |
The intent score is:
I = min(10, Σ(W × F × P))The score lands in daily_intent. The page gets the score in frontmatter. The system now has a receipt for the morning.
What It Loses
The model sees traces. It sees words, commitments, verbs, quantities, and hesitation. It does not see the body that wrote the page.
That loss is the boundary. The system can make recurrence visible. It can show that the same promise keeps returning with a different costume. It can make the morning harder to launder into a feeling of sincerity.
It cannot decide what the page meant. It can only make me face what the page said.
What It Prevents
The Mind Reader prevents one small evasion: writing a promise, closing the notebook, and letting the promise dissolve into mood.
The absurdity is useful because the machine is less sentimental than the ritual. It turns the fog into a list of facts, then leaves me with the debt.