Laya

The Real Problem Isn't Volume - It's Fragmentation The problem with unified inboxes, notification aggregator, and "productivity" tool out there is that they solve the wrong problem. They assume the issue is having too many notifications in too many places, so they put them all in one place. But that just gives you one place with too many notifications. The real problem is threefold: 1. Redundancy. The same event triggers notifications on 3-4 platforms. A Jira ticket update pings Jira, Slack (via integration), email (via Jira notifications), and sometimes the linked PR on Bitbucket. You end up processing the same information four times. 2. Context fragmentation. The same issue, project, or decision is scattered across platforms, each with its own vocabulary. Humans naturally give things different names depending on context -- a colloquial shorthand in Slack, a formal title in Jira, a technical description in a PR. No existing tool connects these. 3. Work-before-work. Before you can make a decision, you need to reconstruct context. That means opening 3-4 tabs, reading threads, cross-referencing ticket IDs, and mentally building a picture of what happened. This is significant overhead every single day. I needed a tool that didn't just collect notifications -- it needed to understand them, connect them, and do the reconstruction work for me. So I built one. ## What Laya Does Laya is a desktop app that sits between your tools and your attention. It intercepts events from Jira, Slack, Gmail, GitHub, Bitbucket, Calendar, Linear, Outlook, and Notion -- then does three things before you ever see a notification: ### 1. Classify and prioritize A fast LLM (the "router") classifies each event: What category is this? (Code, Comms, Ops, Finance, People) Which specialized persona should handle it? (Engineer, Comms, Ops, Sales, HR, Finance) How urgent is it? The router also generates a research plan -- what additional context would be needed to make a decision. ### 2. Connect across platforms This is the part I'm most proud of. Laya builds cross-platform coherence using three layers: - **Explicit links:** Deterministic cross-references from platform data. If a PR description mentions "BUG-1234", that's a direct link. - **Semantic similarity:** ChromaDB vector search finds items that are about the same thing, even if they use different words. The Slack message about "the payment thing" has high semantic similarity to the Jira ticket "NPE in PaymentHandler." - **LLM verification:** For borderline matches (similarity score between 0.20-0.30), an LLM confirms whether two items are actually related. The result: "BUG-1234" in Jira, "the payment thing" in Slack, "fix: null check in payment handler" in Bitbucket, and "status of the payment issue" in Gmail all get linked automatically. ### 3. Stage actions A stronger LLM (the "stager") synthesizes all the connected context and drafts what you'd probably do next: a reply to the email with a status update, a review comment on the PR, a Jira status change. These show up as Action Cards -- you approve, dismiss, or open a workspace for complex tasks. By the time I open Laya in the morning, I don't have a flood of notifications. I have a manageable set of Action Cards, each with the research already done and actions already staged. A few clear decisions instead of a morning lost to context reconstruction. ## Why "Proactive" Is the Key Word There's a distinction that matters here, and it's easy to miss because the AI space is so noisy right now. Most AI tools -- even good ones -- are prompt-based. You open an AI email client and say "summarize this thread." You ask a coding assistant to "review this PR." You invoke an agent with "what's the status of this project?" The intelligence is real, but the first step is always yours: you identify what needs attention, you invoke the tool, you frame the request. Laya eliminates that entire first step. Events flow in from your tools continuously. The classification, context linking, and action staging happen automatically, in the background, as events arrive. By the time you open the app, the work isn't waiting for your prompt -- it's already done. This is the difference between a reactive assistant and a proactive system. You don't tell Laya "summarize the overnight Slack activity." It already has. You don't ask it to "find the PR related to this Jira ticket." It already linked them. The cognitive load Laya enriches daily workflows by providing a unified command center across various tools. It functions as an intelligent email client and integration hub that allows users to connect multiple platforms to manage their tasks proactively rather than reactively. Key Features: - Cross platform notification aggregation and contextual association - Semantic search across all processed notification content - Integration with local agents (claude code, codex, gemini-cli, pi.dev) - MCP server - Multi level summarisation