What the AI does, and what it doesn't

Dayspring uses a language model. You should know exactly for what.

The rule

Facts come from code. The model only selects. Quotes are verbatim.

When the Ascent tells you something ran through your autumn, a program counted that — not a model’s impression. When it shows you a line that carried the season, that line is one you wrote, reproduced exactly.

What the model actually does

• Chooses which of your sentences to show you.
• Groups related things — prayers about the same subject, entries about the same thread.
• Transcribes your voice recordings and your handwriting.
• Interprets a filter you typed in plain English into actual filter settings. It sets the criteria; it never decides which entries match.

What it never does

• It doesn’t write summaries of your spiritual life in its own words.
• It doesn’t tell you what your writing means.
• It doesn’t assess, score, rank or grade anything about you.
• It doesn’t invent a quotation, an event, or a pattern. If it can’t point at something you wrote, it stays quiet.

That last one is the whole discipline. A journal that reflects your life back to you is worthless the moment it starts making things up — and worse than worthless if you believe it.

Your writing and training

Your entries are not used to train models. Not ours, not anyone’s.

Why not more

We could build something that wrote you a warm paragraph every Sunday about how you’re growing. It would be easy, it would feel good, and it would be fiction.

The whole promise of Dayspring is that what you see is what was actually there. Anything invented, however encouraging, breaks it.