Sometimes the most impactful fixes are the ones that restore what should have been working all along. This week, we shipped a fix to Sabine that addresses a frustrating gap in how the agent handles reminder searches during conversations.
The Problem
Users would ask Sabine questions like "Do I have any reminders about the quarterly review?" or "What reminders did I set for the product launch?" — perfectly reasonable queries. But Sabine couldn't find them.
The reminders existed. They were stored correctly. But the search mechanism wasn't connecting topic-based queries to the underlying reminder data. Users had to either list all reminders or remember the exact wording they'd used when creating one — neither of which is a good experience when you're trying to work naturally with an AI partner.
What We Changed
The root cause was in how Sabine's search layer was matching conversational queries against reminder metadata. Topic-specific searches weren't properly mapping to the reminder content and context fields.
We updated the search logic to:
• Expand the searchable fields to include reminder titles, descriptions, and associated tags
• Improve semantic matching so conversational language maps to stored reminder content
• Handle partial matches more gracefully, returning relevant results even when the query doesn't exactly match stored text
Now when you ask Sabine about reminders related to a specific topic, project, or keyword, the agent can actually find them.
Why It Matters
This fix might seem small, but it touches on something fundamental to how people work with AI: trust. When you ask your AI partner a direct question and it can't find information you know exists, that erodes confidence in the entire system.
Reminders are a high-stakes feature. They're how users offload cognitive burden — "remind me to follow up on this next week" or "remind me to review this before the meeting." If the retrieval doesn't work, the whole value proposition breaks down.
With this fix, Sabine becomes more reliable as a working memory partner. You can ask natural questions and get the answers you expect.
What's Next
This reminder search fix is part of a broader effort to improve how Sabine handles context and retrieval across all conversation types. We're working on:
• Enhanced semantic search across all user data (notes, tasks, reminders, and conversation history)
• Better handling of temporal queries ("reminders I set last week," "upcoming reminders")
• Proactive reminder surfacing — Sabine suggesting relevant reminders based on conversation context without being explicitly asked
The goal is simple: make Sabine feel less like a chatbot and more like a partner who actually remembers what you told them.
If you've been frustrated by reminder search in the past, give it another try. And as always, let us know what's working and what isn't — that's how we build better.