AI 2027 × AI 2040 / an independent reading guide
Research snapshotTwo futures.
One unfolding
reality.
One scenario imagines a race to superintelligence. The other asks what it would take to slow down. Put both beside what has actually happened—and follow the evidence for yourself.
Documented pastConditional futures
A map of the narratives, not a forecast model or a probability estimate.
02different scenarios, kept distinct
—documented events & publications
—scenario checkpoints to explore
—latest dated event in this selection
The evidence, month by month
One chronology. Different kinds of claims.
Start in January 2026. Actual dates determine the month; the scenarios alongside keep their own original timing. Icons are original topic illustrations, not photographs of events.
These are not competing countdown clocks. AI 2027 is a fast-takeoff scenario with two endings. AI 2040's Plan A is primarily a policy recommendation: its 2040 endpoint depends on a deliberate slowdown, not merely slower technical progress. Read the method →
Not month-dated: original scenario checkpoints
These original scenario periods are not monthly predictions or facts. Related reports above are dated to their actual publication or documented occurrence, not these scenario years.
Beyond a single line
What changes at the fork?
The original sites explore choices, not just dates. Switch paths to read a short, original summary. These are scenario descriptions, not endorsements, likelihood estimates, or observed events.
Which kind of intervention?
Read the original policy comparison ↗How this comparison is made
Three different kinds of statement.
Scenario: a condensed, original paraphrase of a named section in AI 2027 or AI 2040. Fictional companies such as OpenBrain and DeepCent are not treated as real organizations. Where one scenario has no matching year, its actual narrative period is shown.
Observation: a dated release, publication, legal change, announcement, or incident with a direct source. A report being published is an event; the report's projections are not observed outcomes. Company benchmarks remain company-reported. Investment commitments are not completed construction.
Mapping: our editorial explanation of how that observation bears on a particular scenario mechanism. A precursor is not confirmation of an endpoint. No automated truth score, probability update, or percentage-complete metric is assigned.
Read the counterevidence, too.
METR's early-2025 developer slowdown is explicitly historical and paired with its February 2026 update. Stanford's employment results are descriptive, not a causal estimate. AISI's incident involved deliberately permissive evaluation settings; it is not presented as a sandbox escape or normal consumer behavior.
Daily actuals checks, not invented daily news.
Content snapshot dated loading. The latest dated event included is loading. A daily check can find no qualifying additions; it does not redate existing events or prove nothing happened. This remains a curated record, not a complete news archive.
Both site timelines use a single daily workflow at 7:00 am Sydney time. New companion entries require their own reviewed scenario connection and verified original source. The published check status and content-change time are separate; failed runs retain the last good record. An overdue check is labelled rather than silently called fresh. The original scenario text and timing are not automatically rewritten.
Event dates and source publication dates are separate. Revisions and effective dates are identified. Month-only dates remain month-only and are grouped within their month, not assigned a fabricated day. An undated policy page is labelled as such. Opening this page does not refresh the research.
Sources were read through public web retrieval or a browser. Those access checks are not endorsements or independent replication. The selected source set includes multiple reports from the same organization; repeated references to the same source or incident are not independent corroboration. Some publishers may restrict access in your location.
Attribution & scope.
Original scenarios: AI Futures Project's AI 2027 team and AI 2040 team, with original websites by Lightcone Infrastructure. This independent page is not affiliated with or endorsed by them. Summaries, diagrams, and mappings here are original; consult the originals for their complete arguments and updates.
This page does not modify Peter Xing's forecast or the main site's evidence accounting. Daily operator-side checks reuse the existing source-verification process; opening this page does not collect news. Filters stay in the URL; there are no accounts, analytics, background requests, or browser-storage records.