40,000 screenshots a day. One page each morning.
Daybrief keeps only the captures that show something new, then writes one report per person from what is left.
The cut is the whole point. Sending every capture costs $74.38 a day, or $1636 a month, for the same reports.
Yesterday, in three reports.
A witness day at Harbor Legal Staffing, three people in two roles. Minutes come from the activity log. Sentences come from the captures, and every one of them names the captures it rests on.
Drafting the Alvarez motion to compel and its exhibit index
Maya spent most of the day drafting and revising the Motion to Compel in Alvarez v. Northwind Freight, moving between the Statement of Facts, Meet and Confer, Argument, and Relief Requested sections.
She researched supporting law on LexBrief, first on the meet and confer requirement and later on the discovery sanctions standard.
She checked the Alvarez matter record in CaseHarbor repeatedly, including the Requests for Production, Defendant responses, and a note that the custodian list arrived without a description of recordkeeping systems.
In the litigation-support Loop channel she coordinated with Dana Vogel and Owen Baptiste on the hearing date change and the order of sections in the motion.
Later she started building the Exhibit Index for Alvarez and replied to Karen Whitlock and Ingrid Marsh in her inbox about the dispatch records and a new Tolliver intake.
Not visible in the captures. Time spent on any phone calls or in-person meet and confer discussions with opposing counsel is not visible. The lock screen period around midday leaves that time unaccounted for.
Same role, same day
The two paralegals, side by side.Where 93.3 percent of the pictures go.
The app and URL stream splits the day into 34 segments. The first capture of a segment is never dropped, because the log can name why the screen changed.
Each capture becomes 144 bits. Anything within 6 bits of the last capture kept is the same screen with the cursor somewhere else. 1,440 become 373.
One capture per 15 minutes of a segment, 6 at most, spread across it. 373 become 97.
The threshold was not picked by taste. On this day, captures of the same screen never differed by more than six bits, and real screen changes started at eight, so the cut sits at 6. Change the captures and the measurement is run again.