CantSleep

Inspiration

• I couldn't sleep last night.

What it does

• It helps users fall asleep by identifying what's keeping them up and then delegating to specific agents that help handle those cases. ## How we built it
• I prompt-engineered an agent that didn't feel overwhelming or obviously "agent-y" and felt more lax to ensure a better user experience.
• We didn't build them out in the timeframe, but ideally, we'd have agents that have access to the tools they need for solving the specific user issue. -- For example, if the user is stressed, we could potentially connect to their Google Home and play mind-soothing music in the background, as our agent gently guides them away from their stress.

Challenges we ran into

• getting consistently "chill" responses from our agent who didn't feel like a retail customer service agent.

Accomplishments that we're proud of

• Ideation of potential tools that could address specific user sleep issues

What we learned

• Balancing functionality with empathy in AI interactions.
• The complexity of tailoring AI responses to be personal yet effective.
• Insight into the diverse factors affecting sleep and how an AI agent can address them.

What's next for CantSleep

• building out subagents that handle specific tasks
• connecting with other devices to give the user a strong and real experience

Built With

• googlehome
• smartthings
• spotify
• vertex