AI email is stuck on the wrong surface

A list can only do two things

An inbox is a list. Newest at the top. That was a reasonable design in 1993 and it is still the design today.

When you put AI on a list, there are exactly two moves available. It can make the list shorter, or it can reorder it. Summaries make items shorter. Priority scores move them up and down. That's the whole space.

So you get a smarter list. You still open the thread, still switch to the calendar tab to check Thursday, still copy the deadline into a task app, still come back and do it again with the next one. The tab switching is where the day goes, and no summary button touches it.

Give the work a place to happen

We rebuilt the inbox as an infinite board. Mail, calendar, tasks, documents, and agents sit next to each other as frames on one canvas. Nothing is nested. Nothing is behind a tab.

Now the two moves become a lot of moves.

Drop an email on Thursday and it becomes a meeting, because the calendar is right there and the agent can see both. Sweep sixty threads into a pile that keeps itself sorted. Pull the promise out of a thread into a todo that sits next to the mail it came from. Wire an inbox into an agent that runs every morning and hands you what needs answering.

Same model. Different surface. The difference is that the agent now has somewhere to put things, and you can see where it put them.

Watch it work

Most AI in software is a black box with a spinner. You type, it thinks, something happens somewhere.

Ours works in the open. Harry is the agent you talk to. He reads your board, drafts in your voice, schedules, and keeps things in order. Other agents join him on the paid plans, each with one job and a name.

When Harry files sixty threads, you watch the piles form. When he drafts, the draft appears as a dashed ghost right where it belongs, and you apply it or you don't. When he finds time for a meeting, the hold shows up on your actual week, dashed, waiting for you.

You are not reading a report about what happened. You are looking at it.

The last click is yours

Here is the part we care about most.

No draft an agent writes can leave without you pressing send. Not as a setting. Not as a policy we promise to follow. Every AI-written draft lands in one outbox, and there is no other path out of the product. We built it that way on purpose, and we are not going to build the other path later.

Anything else the agent can't reverse gets staged for your approval instead of done. A calendar hold is dashed until you apply it. A todo is a ghost until you keep it.

The whole category is asking you to trust an agent with your outbound mail. We would rather not ask. We would rather make it structurally impossible for the question to come up.

Where this goes

We think the interesting work in AI and email isn't a better summarizer. It's giving the agent a surface with more than one dimension, and giving you a way to see what it's doing on that surface.

That's the board. Your whole day, laid out the way you think.

A note from us

We named this thing after a messenger owl, which tells you roughly how seriously we take ourselves.

Hedwig started because we were both losing hours a day to the same small indignity: reading something, understanding it in two seconds, and then spending four minutes moving it around between four apps that don't know about each other. We tried the plugins. We tried the smarter inbox. It never got at the actual problem, which is that your day is spread flat across tools that can't see one another.

So we built the surface we wanted. It began as an email client and grew into a board, mostly because every time we added a piece, the honest place to put it was next to the mail rather than behind it.

We are a very small team with a very long list. Some of what you use today was built the week before you saw it. Some of it will be better in a month because you told us it was rough. That is not a disclaimer, it's the plan. The people using Hedwig right now are shaping it more than any roadmap we've written.

Two things we'd ask of you.

Tell us when it breaks. There's a community forum and a support address, and both go to people who can actually change the code. We read everything. We answer more of it than is probably reasonable for a company our size, and we would like to keep it that way for as long as we can.

And hold us to the outbox. It is the one promise in this product we consider load-bearing. Your mail is yours. Your voice is yours. The agent works for you and stops at the send button, every time, forever. If we ever drift from that, say so loudly, and we will fix it.

Thanks for being here early. It means more than it probably looks like from the outside.

— Nesh and Neil