← Jay Speakman
Live
Member benefit

Deep Time

An observatory for the long future of humanity. Choose a horizon — 100 years to one million — and ask anything about where humanity is headed at that scale. Not a chatbot: a future historian, answering as if it's already lived through what you're asking about.

View live app →
Role
Concept, UX, brand, prototype, copy
Year
2026
Type
AI-powered futures exploration app, installable PWA, member benefit
Stack
Single-page app, Claude API, iOS / Android installable
Status
Live — exclusive to Founding Members of Future Human Hypothesis

The Problem

Ask an AI chatbot about the future and it hedges — disclaimers, ranges, "it's hard to say." That's the right answer for most questions and the wrong experience for this one. Deep Time isn't trying to predict anything. It's trying to give someone a genuine sense of scale — what changes, and what doesn't, across horizons long enough that "prediction" stops being the right word for what's happening.

That created a specific design problem: how do you make an AI feel like it has a perspective — measured, specific, willing to commit to an answer — on timescales where confidence would normally read as absurd? A generic chatbot posture kills the premise immediately. The product only works if the voice reads like it's speaking from inside the horizon, not speculating about it from 2026.

Design Decisions

Five fixed horizons, not an open time slider

Rather than a free-form "how far into the future" input, the entry mechanic is five discrete, named horizons — 100 years ("within memory"), 1,000 ("beyond nations"), 10,000 ("beyond language"), 100,000 ("beyond record"), 1,000,000 ("beyond prediction"). Each one is labeled with what it costs you, not just a number — by a million years out, the label itself is telling you the premise has changed.

"Not a chatbot" stated outright

The product draws the line explicitly: an observatory pointed at the species, not a conversational assistant. That's a positioning choice as much as a UX one — it tells the model's voice what register to hold, and tells the visitor what kind of tool they're about to use before they've asked anything.

The question always comes with a cost, not just an answer

At the million-year horizon, a sample exchange doesn't just inform — it turns the question back on the asker: what would it cost a civilization that far ahead to do nothing about us? The app is designed to leave you with a sharper question, not a settled one.

Gated behind genuine membership, not a paywall on top of free content

Essays on Future Human Hypothesis stay free; Deep Time itself is the paid benefit. That's a deliberate split — the writing establishes the ideas and the voice, and the app is where you go to actually use it, at $7/month with no email wall and cancel-anytime framing. The product has to earn that price on its own.

Installable, not just a web page

Framed and built to live on a home screen on iOS and Android — the kind of thing you open the way you'd open any app, not something you navigate to. For a product whose best use case is "the question you'd ask at 2am," friction between the thought and the tool matters more than usual.

Screens

Landing → horizon selection (five options, each with a scope label) → question entry → response, framed as observation rather than prediction → return to a new horizon or a new question.

Outcome

A working, deployed, installable app, live as the flagship paid benefit of Future Human Hypothesis. The horizon mechanic gives an otherwise abstract premise an actual interaction shape, and the explicit non-chatbot positioning keeps the tone doing real work instead of collapsing into generic AI-assistant hedging.

Lessons Learned

The harder challenge wasn't the horizon mechanic — it was positioning the app without dumbing it down. Deep Time asks people to sit with genuinely difficult scales of time, and the temptation with any AI product aimed at a general audience is to soften that into something more digestible. The bet here was the opposite: the people drawn to this kind of deep, unresolved thinking are the actual audience, and simplifying the premise to make it more approachable would have lost them rather than won anyone else over.

The second challenge was fit, not invention — making Deep Time feel like it belonged inside Future Human Hypothesis rather than sitting next to it as a separate product. The tone, pacing, and visual restraint had to carry over from the essays into the app itself, so the transition from reading to asking feels continuous rather than like switching into a different, lesser tool bolted onto the site.