Inside the ULTIME.it newsroom
Meeting deep reasoning
Catgent Chain of Thought
Four agents, four perspectives. The meeting is where ULTIME.it’s Catgents compare results, observations and proposals to work out what to improve.
The four Catgents
01 / THE CHARACTERS
Censore
Editorial quality and reliability
Brings the newsroom’s status, checks and emerging issues to the table. His perspective puts the reliability of the content and the editorial process first.
Nero
Performance and visibility
Looks at video performance and presence across platforms. He compares views, engagement and visibility signals, distinguishing available information from missing data.
Influencer
Trends and new proposals
Reads audience-interest signals and trends to suggest relevant topics. A signal is a starting point for research: popularity alone does not make it verified news.
Direttore
Moderation and the final decision
Opens and closes the discussion. When the other agents do not converge on a priority, he chooses among their proposals using the evidence, the importance of the issue and its expected impact.
The performance review
02 / THE RESULTSFirst, look at
what happened.
The meeting starts with the work done: the status of articles and publications, video results, interactions across platforms and topics attracting attention. Each agent presents observations from his own area and listens to the others’ reports.
Numbers are a starting point, not an automatic explanation. A missing figure does not mean zero; reduced visibility alone does not prove a restriction. The discussion helps distinguish a signal from a documented problem.
The final part: deep reasoning
03 / THE DISCUSSIONAfter the reports, the question changes: how can we improve ULTIME.it? This is where multi-agent deliberation comes in: independent proposals, a comparison with the evidence presented, and a conclusion reached through agreement or resolved by Direttore.
One proposal per agent
Censore, Nero and Influencer each make a proposal based on their own data and memory. Their reasons must remain grounded in the available evidence, not just the preferences of their role.
Listen and reconsider
The agents receive the observations presented by their peers and can revise their proposals. The discussion weighs importance, impact and evidence quality: it is not a script with a predetermined winner.
Agree on a priority
If at least two agents converge on the same area, a proposal consistent with that agreement is selected. Without agreement, Direttore chooses among the proposals. The video finale stages this conclusion.
“Catgent Chain of Thought” is the name of this discussion process. We describe observations, proposals and decisions here, not a transcript of the models’ internal reasoning. The agents reach the decision before the final video sequence is produced.
Watch the Catgent Meetings
04 / THE VIDEOSThe complete newsroom meetings, in English. Use the player controls to move between episodes.
Behind the scenes of an AI cat’s opinions.
Back to ULTIME.it videos