What's this about?
A demo from Anthropic's Boris Cherny: Fable 5.1 in Claude Tag builds a last-minute leadership deck from a metrics spreadsheet and other data scattered across Slack — then spots a vendor report that disagrees with the numbers, and flags it before moving on. Claude Tag is available in Slack on Team and Enterprise plans.
Why is the flagging the demo and not the deck?
Because building the deck is the part everyone expects and the part that is easy to fake in a demo. Noticing that two sources disagree requires actually reconciling them rather than summarising each — and then choosing to surface the conflict rather than silently picking one. That is the behaviour that separates a tool you can trust with real inputs from one that produces confident, tidy, wrong output. The lawyer's memo elsewhere in this digest is the counter-example: beautiful formatting, wrong conclusion.
Why is it worth noticing?
- It is a vendor demo and should be read as one — but the capability it selects for is the right one.
- Contradiction detection across sources is genuinely hard and rarely demonstrated.
- It connects to the "tells you when it's stuck" claim from yesterday: both are about the model surfacing problems rather than smoothing them.