Every AI tool you use today is reactive. You ask a question, it answers. You give a prompt, it generates. The tool never says, "Hey, I noticed something you should look at." It never volunteers work. It just waits.
Today we're changing that with Frost Dispatch.
From Reactive to Proactive
Frost Dispatch is the natural evolution of Frost, SLEDS' AI layer. Until now, Frost analyzed your sled's context to generate Pulse briefings — narrative summaries of what's happening across your threads and assets. Dispatch takes it further: Frost now proposes work.
Here's what that looks like. Frost notices that your architecture doc references an auth flow that was replaced three weeks ago. Instead of silently knowing this, Frost creates a dispatch: "Update architecture overview — the doc still references the old auth flow. Based on the Clerk migration thread, I can update it to reflect the current implementation." You see the proposal, approve it, and Frost handles it.
How Dispatches Work
Every dispatch has a clear anatomy: a title, reasoning, source references, context bundle, and priority. Frost explains why the work matters and what context informed the proposal. You can approve it for Frost to handle directly, queue it for pickup in Claude or ChatGPT, or dismiss it with a reason.
Dispatches can be scoped to a sled (like updating a shared spec) or personal (like prepping for a 1:1 by pulling context from relevant threads, emails, and calendar events). There are four autonomy levels — from full manual approval to fully autonomous — so you control exactly how much Frost can do without asking.
The Dispatch Feed
Your SLEDS homepage now includes a dispatch feed showing what Frost has proposed, what's in progress, and what's been completed. It's like a to-do list that writes itself, grounded in the actual state of your projects rather than aspirational planning.
Early testers told us the most surprising part isn't the individual dispatches — it's the pattern. Frost catches things that fall through the cracks: specs that reference outdated decisions, conflicts between conversations in different tools, prep work for upcoming meetings. These are the tasks that don't belong to anyone until they cause a problem.
This Is What Shared Context Enables
Frost Dispatch is only possible because of the shared context layer underneath it. An AI assistant that only sees one conversation can't detect cross-thread conflicts. It can't notice that an asset is outdated based on a decision made in a different tool by a different person. Shared memory makes proactive intelligence possible.
We're starting with Level 1 — full manual approval for everything. Over time, as teams build trust with Frost, they can dial up autonomy. The goal isn't to replace your judgment. It's to surface the work that needs judgment, before it becomes a problem.