Skip to content

The Ideas Behind Spillway

Spillway is not just an email program with AI added to it.

It is built around a view of how people actually deal with information, attention, uncertainty, relationships, and decisions over time.

That perspective comes from communication research, political science, psychology, data science, and human-computer interaction as much as it comes from software engineering.

Email is a social system

An inbox is not simply a pile of documents.

It is a record of people asking, promising, reminding, coordinating, persuading, sharing, and changing their minds. The meaning of a message depends on who sent it, what came before it, what project it belongs to, and what happened afterward.

Spillway therefore treats messages as part of changing human systems rather than as isolated pieces of text.

Evidence has a source

A suggestion should never appear as if it came from nowhere.

Spillway is designed to keep track of where a belief or suggestion came from: a message, a calendar event, a human correction, a past pattern, or another observation. This is sometimes called provenance, which simply means knowing the origin of the evidence.

That matters because people should be able to ask:

  • Why does Spillway think this is a task?
  • Which message supports this conclusion?
  • Did a person confirm it, or did an AI model suggest it?
  • Is the supporting information recent or old?

Confidence should change over time

Human systems do not stay still.

A plan may be likely today and abandoned next month. A weak clue may become convincing after several later messages. A once-useful pattern may stop being reliable when a person's role, project, or habits change.

Spillway is designed so confidence can rise, fall, and be revised as new evidence arrives.

The project draws inspiration from Bayesian reasoning: the idea that beliefs should be updated when new evidence appears. In practice, Spillway uses a more flexible, quasi-Bayesian approach. That means it follows the same basic principle without pretending that every human signal can be reduced to one perfect probability formula.

Older evidence may matter less

Some facts remain important for years. Others become stale quickly.

Spillway therefore needs to distinguish between preserving history and giving old information the same weight forever.

This is where decay matters. Decay means that some evidence becomes less influential as it ages unless newer information reinforces it. A tentative plan from two years ago should not automatically outweigh a clear update from last week.

Decay should be explainable and appropriate to the kind of information involved. It should never mean silently deleting history or erasing a human decision.

Several clues can be stronger than one

One AI response should not become the truth.

Spillway is designed to combine multiple observations over time. A request may be supported by the wording of a message, a later reply, a deadline, a repeated pattern, and a human confirmation.

The system should become more confident when independent evidence points in the same direction and less confident when the evidence conflicts.

Human decisions are part of the evidence

A correction is not merely an override button.

When a person confirms, rejects, or changes a suggestion, that action provides valuable evidence about meaning and intent. Spillway should learn from those decisions while still allowing them to be revised later.

This is why human judgment remains authoritative without being treated as frozen forever.

Attention is a limited resource

The product is also shaped by research on attention and task switching.

People pay a cost every time they shift between unrelated kinds of work. An inbox that mixes student questions, family scheduling, receipts, newsletters, documents, and deadlines forces repeated mental resets.

Spillway aims to reduce those switches by helping people review similar work together and return to information when they are ready to act on it.

Why these fields belong together

Communication research asks how information moves among people and how context changes meaning.

Political science studies institutions, incentives, uncertainty, collective behavior, and decision-making with incomplete information.

Psychology examines attention, memory, judgment, forgetting, and how people revise beliefs.

Data science helps organize large, messy collections of information and find useful patterns without assuming the data are perfect.

Human-computer interaction asks how software can support people without confusing, overwhelming, or quietly taking control from them.

Spillway sits naturally at the intersection of those questions.

Why this project comes from a communication researcher and political scientist

Josh Pasek's work spans communication, political science, public opinion, survey methodology, emerging technology, and data science.

Those fields all deal with a common problem: how people make sense of incomplete, changing, and socially produced information.

That background does not replace software engineering. It shapes the questions the software is built to answer:

  • What counts as evidence?
  • Where did that evidence come from?
  • How strong is it?
  • How should newer information change an earlier conclusion?
  • When should old information matter less?
  • How do relationships and institutions change meaning?
  • When must a human remain in control?

The reason Josh wanted Spillway and the reason this academic background is useful are related, but not identical.

He wanted the product because ordinary email was consuming too much attention and hiding too much work. His research background helped provide a deeper model for how the solution should reason about people, evidence, uncertainty, and change.

Not just clever code

Spillway uses software engineering and AI, but its goal is not to show off either one.

The deeper aim is to build a system that behaves sensibly in dynamic human environments: one that can explain what it believes, revise itself when the world changes, preserve the source of its conclusions, and leave important decisions in human hands.

Read why Spillway exists · Learn how Spillway is built · Read the development philosophy