Fatima Ashraf

Huge congrats to the Markopolo team on this amazing shift! My question centers on the LLM's role: How does the system ensure conversational coherence and brand consistency across distinct channels (e.g., SMS, Email, Voice) while simultaneously adapting the message in real-time based on the latest signal?

Given that you're now incorporating sophisticated features like the Real-Time Stress Relief and AI Coach, all of which demand high data fidelity, my question is: How are you managing the signal-to-noise ratio and data normalization across such a massive and disparate number of third-party sources (from Garmin to Reddit) to ensure the AI's recommendations remain reliably accurate? Upvoted!

Fascinating application of multimodal AI. The technical challenge of generating coherent, context-aware visual sequences from unstructured audio, rather than just abstract visualizations is immense. How effectively is the model able to handle complex thematic shifts and distinguish between internal emotional context versus external descriptive context in the speech? This could redefine...

How does Juno's architecture manage data integrity and performance for users who have 10,000+ completed tasks and a year's worth of time-blocks? Specifically, is there a simple archive/purge mechanism for completed tasks that keeps the active planner "calm" and snappy, while still retaining the historical data for reflection and progress tracking?

For users who rely on the 'Projects' feature (with its dedicated files and custom instructions) for a single client or goal, will Group Chats eventually be allowed to tap into the Project's memory/context, rather than the User's personal memory? Having the ability for a group to converse with an AI that's already trained on the Project's specific documents and goals would turn this from a great...

Are there plans to leverage your new 'Agent' infrastructure to intelligently infer or predict speaker identity based on the context of the Notion page itself (e.g., who's tagged as the meeting owner, who's mentioned in the agenda, or who has permission access) to at least provide a probabilistic label, rather than requiring a dedicated audio/video API integration?

HUGE congrats on the TRAE SOLO launch! This is the launch that definitively moves us from AI co-pilots to true AI engineers. The Plan Mode is brilliant for aligning before any code is written. I can't wait to delegate an entire feature and just focus on reviewing the plan and the diff. Question for the team: How are you thinking about managing the context window size with so many tools (IDE,...

There was a feature in Duolingo where we could host shows. I don't find it anymore. Can anyone tell me if Duolingo discontinued it later on? I was really looking forward to it after completing French B-1