On Groupliv, people who want to buy a new apartment form a group, and developers send that group private offers. The members compare them and vote. When an offer arrives, Liv, the product’s AI advisor, writes one review into the group’s feed: is it cheap or expensive for the area, and what is weak about it. Every member who opens it reads the same review. [1]

One mistake there is worse than all the others: a false sentence about money. Writing “after the discount” about an offer with no discount misleads a whole group about the thing it is negotiating. Telling the model not to was the obvious fix. It held on one run and broke on the next, from the same facts. [2]

Figure: Example. The offer, the wording and the numbers are invented. The steps and the rule (one retry, then drop, only for an offer with no discount and a price per m² computed in code) follow the code that runs in production. See the animated figure in the original note.

The calls I made

The AI writes sentences; code does the sums.…