The future deserves
a closer reading.

This is a public beta

The library is open so that readers with the link can explore it and contribute while it is still being built. Entries may be incomplete or unverified, features are missing, and the design will keep changing. We will try to keep your account, ratings, votes, and suggestions on a best-effort basis, but some or all of them may be reset before the full launch. If something looks wrong, use the report controls on an entry or suggest a change.

A library, not a prediction

A scenario’s inclusion means that someone has described it. It does not establish that the account is correct, probable, or endorsed by the library. We link to original works and summarize their mechanisms and premises so you can inspect the argument yourself.

The focus is the mechanism of change: the path from today to a new situation, and AI’s role at each step. Each step can be classified by who acts (a human, an AI, both, or no one in particular), whether the outcome is intended, and, for AI actors, how autonomous and how aligned they are. The title and source link identify the work. Later plot is secondary, and missing steps in the transition stay visible.

Many entries are threat models: accounts of how advanced AI could lead to catastrophe, set out as a causal pathway with each step and premise in view. Hopeful and ordinary futures sit beside them on the same terms. Comparing the two shows which premises or decisions separate a good outcome from a bad one.

Seven questions, kept apart

Three questions are about the account itself. Specificity / clarity asks how explicitly the original explains the transition. A detailed account can still be implausible, and a compelling world can leave its origins unexplained. Plausibility asks whether the premises and causal chain hold together. Likelihood asks whether a broadly similar outcome will happen this way within the stated time window, and is offered only for entries that state one. A believable mechanism is often one of many, so plausible does not mean likely.

Three questions describe the future the work arrives at, not the path there. Human well-being and machine-mind well-being ask how well off each kind of mind is, with a separate answer for futures where none remain or none exist. Open future asks how much humanity can still act on its circumstances, change, and evolve, whether or not the result so far is a good one. Keeping these apart lets a reader say that a future is comfortable and closed, or that one population thrives while another is gone, without averaging the two away. Plausibility and human well-being together serve as the library’s assessment of how likely a path is and how severe its outcome would be.

The last question, would you want to live in this future, is explicitly personal. It runs from dread to longing with the world today in the middle, and a divided answer is a finding, not noise. The up/down vote is different again: it asks whether the entry is a useful, well-articulated scenario to examine.

Every reader has equal weight. Averages show response counts. Rating-based rankings require at least three responses for that measure; entries below the threshold appear under “Needs ratings.” Unrated entries are excluded when a minimum rating is selected. Leaving a question unrated means insufficient information and adds nothing to the average.

“Top this month” counts each reader’s current vote if it was first cast or changed during the current UTC calendar month. “All-time top” counts all current votes. These are votes of confidence, not measured probabilities.

Premises describe the story

The shared premise list keeps similar ideas comparable. Each entry records whether a premise is explicitly or implicitly true or false in the story, or does not apply to it. Explicit means the source discusses the premise, whether by stating, depicting, or arguing for it. Implied means the source is silent, but the scenario only works if the premise holds, or fails to hold. This records what the story relies on, not what its contributor personally believes.

Instrumental convergence means different ultimate goals can lead to similar instrumental strategies. It does not mean all intelligences converge on the same ultimate values.

Curated contributions

You can submit your own work or someone else’s, with attribution and a link to the original. New submissions enter curator review. The submitter stewards the entry once published; site-seeded entries are stewarded by the library.

Anyone signed in can suggest a structured improvement. The steward reviews the proposed change, explains the decision, and preserves contributor credit and revision history. Suggestions cannot overwrite a newer revision.

Verified original authors can request stewardship. Site admins may revoke stewardship for unreliability or inappropriate conduct, including an author’s stewardship. Authorship verification and those administrative workflows will be added before public contributions open.

Source checks aren’t truth certificates

Source unchecked means the representation is not yet fully checked. Source checked means a curator compared the entry with the original work. Future author verification will record an author’s confirmation of a specific revision. None establishes predictive accuracy.

The 36 founding entries are marked AI-generated draft: their summaries, mechanism steps, and premises were drafted by AI from the original works and public summaries, then reviewed only lightly. They will be replaced with human-edited versions. An entry a contributor wrote with AI help is marked AI-assisted. None of the founding entries has fabricated votes or ratings, and complete source checks are required before public publication.