Every figure below is computed in SQL, not by the model. DeepSeek only writes the interpretation. It is good at spotting what went quiet and which segment converts. It cannot see how a meeting felt, so treat the recommendations as prompts to think, not instructions.
Open recommendations
0
Nothing open
The next weekly run posts here on Monday morning.
Past runs
No runs yet
Each run stores the figures it was given alongside the write-up, so a bad read can be traced to either the numbers or the interpretation.