01The problem
At the end of the day the work has been done, but its trail is scattered: conversations with the AI, commits, changes to the ad accounts, call transcripts, routine outcomes. Rebuilding it by hand takes time; not rebuilding it means starting the next day without a clear picture.
02How it is used
- You ask for the summary with one sentence, when you need it.
- The page opens with a recap: one sentence and four numbers.
- Below, the details in order: the things to look at, a card for each client, the rhythm of the day hour by hour, the outcome of the routines.
- The details stay collapsed and open only when needed. A lens shows what comes from the data and what from the model.
03How it is built
- Collectiona script reads six logs: AI chats, commits, Meta and Google Ads changes, calls, routines
- Factstimes, counts and changes are extracted by code
- Judgementa model writes the summary sentence and chooses what to highlight
- Transparencythe lens colours the code's facts and the model's judgement differently
- Outputa page to read and a copy as a note in the second brain
04Results
The first version was a long note, dropped after a day's trial because nobody read it. The current version reads in thirty seconds and leaves the rest one click away.