Frequently Asked Questions
This page was written before anyone had actually started asking questions. These are the questions we think people may ask. Once real questions arrive, they will be allowed to compete with our guesses.
What is Spillway?
Spillway is a privacy-first email and work organizer for Mac and iPhone. It helps people sort messages, find requests and dates, and keep important work from disappearing into an overflowing inbox.
Why build another email client?
Because the problem is not only reading and sending email. Email has become a place where tasks, documents, schedules, decisions, and personal logistics all arrive mixed together.
Is Spillway ready to use?
Spillway is currently available through a private TestFlight beta. It is still early software, and testers should keep Gmail or another established email client available rather than relying on Spillway as their only way to reach important mail.
See current progress or learn about the private beta.
Does Spillway send my email to its own servers?
No. Spillway does not operate a server that receives and stores users' Gmail content.
Local AI can run on the user's Mac. When a user chooses a cloud AI provider, selected content is sent directly to that provider under its terms.
Can I use Spillway without AI?
Yes, although much of Spillway's distinctive value comes from inference.
AI can be disabled for privacy, policy, safety, or failure-recovery reasons. People who want AI assistance without sending content to a cloud provider can use a local model.
Do I have to use one particular local AI model?
No. Spillway treats the model as replaceable.
Different local models trade capability, speed, memory use, and context length.
The local-AI guide explains the practical tradeoffs.
Doesn't local AI know much less than cloud AI?
Often it knows less about the general public world.
But neither local nor cloud models automatically know your private working world: current projects, relationships, rosters, prior corrections, or months of private mail they have never seen.
Spillway is designed to provide relevant context when appropriate rather than assuming the base model already knows it.
Can knowledge learned from one account help with another account that cannot use AI?
Potentially, when policy permits and useful knowledge can be represented safely without moving restricted source content.
That is part of the intended architecture, not a claim that every form of cross-account learning is available today.
Read about AI and privacy boundaries.
Can I choose different privacy rules for different accounts?
That is part of the design.
Different mailboxes may reasonably use local AI, cloud AI, or no AI. A combined workspace should not erase those boundaries.
Why does Spillway ask me to review AI suggestions?
Because AI is useful without being the final authority.
Spillway treats model output as a proposal that a person can confirm, correct, reject, or ignore.
Read how Spillway separates evidence, inference, and decisions.
Will Spillway learn how I organize email?
That is the goal.
Human decisions can become evidence for later suggestions while remaining correctable and attributable to their source.
Read the ideas behind Spillway.
What is Background Learning?
Background Learning is Spillway's way of revisiting older information in bounded work so useful context does not have to be rediscovered from scratch every time.
It is designed to remain subordinate to interactive use of the computer rather than consuming resources simply because they are available.
Why is a communication researcher and political scientist building an email client?
Because email is also a social-science problem: people, institutions, relationships, incomplete information, attention, memory, and decisions all shape what a message means and what should happen next.
Read the social and statistical science behind Spillway.
Why publish so much about Spillway if the code is private?
Trust does not require publishing every implementation detail.
Spillway publishes product philosophy, architecture explanations, scientific reasoning, privacy commitments, design tradeoffs, and development methodology so people can inspect the ideas and assumptions behind the software.
Public Ideas, Private Code · Documentation provenance
Will all my emails turn into AI slop?
No. The goal is to reduce clutter and recover useful information, not replace every message with a longer, more enthusiastic version of itself.
Don't all professors hate AI?
Professors are unusually skilled at disagreeing with one another.
Is this app going to speed up the AI takeover?
That would be an unfortunate product-market fit. If the machines eventually take over anyway, we hope they at least respect the category hierarchy.
Does Spillway know what I am thinking?
No. It sometimes has enough trouble determining whether a newsletter is actually a newsletter.
Can Spillway finally make me answer my email?
We admire your confidence in software.
Will this get me to Inbox Zero?
Spillway is aiming for something more durable: an inbox that is understandable and under control.
Why are there so many words on this website?
Because the project is trying to explain how the software works before asking people to trust it.
Also, the person building it is a professor.
Did we miss your question?
Almost certainly. Questions, feedback, and beta inquiries may be sent to software@joshpasek.com.
Documentation provenance: Iterative human–AI construction. See Documentation Provenance.