Scenario, not forecast. Every number is an assumption, and every lab, model and agency named here is fictional. As of 25 September 2026.

RISCON LABS · SCENARIO & AUDIT

Assume that in 2030, ASI makes the judgments. Learn to question them now.

Follow, quarter by quarter, two worlds given the same ASI, and practise auditing with the judgment records they leave behind. An auditor need not be smarter than the ASI, only able to reproduce its results and check that its conditions held.

The central question

When ASI makes the judgments in 2030, what must I ask to be able to correct them?

  • 25quarters
  • 8axes
  • 236judgment records
  • 7auditor levels
Read the scenarioAuditor training

RISCON’s two axes

Co-evolution and Audit

Axis 1

Co-evolution

People ask; ASI judges and corrects. Carry Meridian’s question loop into 2030 and it becomes five steps.

  1. Finding the questionPeople
  2. JudgmentASI
  3. The audit questionPeople
  4. CorrectionASI
  5. ReviewTogether

In today’s Meridian, people make the judgments. ASI 2030 begins after judgment has passed to ASI; what people must then keep is the right to ask and the ability to correct.

Meridian

Axis 2

Audit

Draw the ASI society out to both extremes as scenarios, train hard at questioning ASI, and become an ASI auditor. ASI may come to do nearly all cognitive work better, but asking questions from the position of the person being judged cannot be handed over to the side that judges. That is why auditor training is almost the only vocational training left.

  1. Scenarios at both extremes
  2. The ability to ask
  3. The ASI auditor

The training for ASI auditors is free and non-commercial.

RISCON Labs · added 2026-11-10 · why one auditor, and why now

Each Person Faces the Answer Alone

When answers pass to ASI, each person will stand before those answers alone. A litigant without a lawyer does. So does a patient who has just been given a diagnosis. No committee or evaluation team sits beside them. So the unit of training has to be one person, and the teaching material has to be real decisions whose outcomes are known, not opinions. This is hard to see today because AI is still visibly wrong. Once most answers are right, the ability to find the rare flaw alone will be the scarcest skill, and by then it will be too late to begin.

A working session on 10 November 2026 is one such case. A model was given this premise and read it as something else, and one auditor caught it.

The judgment records on this site are scenarios. Audits against real outcomes are run separately by RISCON Labs.

The fork · 2028 Q4

One Technology, Two Worlds

In both endings the ASI, P-6, is the same system. What differs is whether people kept a place to ask. The fork is the forty-eight hours of the power-grid crisis in the winter of 2028.

T1 · 2026 Q4 — Unread ApprovalsT2 · 2027 Q1 — The Autonomous EngineerT3 · 2027 Q2 — The RepricingT4 · 2027 Q3 — The First SealT5 · 2027 Q4 — Six YesesT6 · 2028 Q1 — The Autonomous ResearcherT7 · 2028 Q2 — The First AuditorsT8 · 2028 Q3 — The Great ReassignmentT9 · 2028 Q4 — Forty-Eight HoursI1 · 2029 Q1 — The Judgment Audit ActsI2 · 2029 Q2 — Cross-AuditI3 · 2029 Q3 — Eleven DaysI4 · 2029 Q4 — The Price of SpeedI5 · 2030 Q1 — The Audit AccordI6 · 2030 Q2 — An Audited ASII7 · 2030 Q3 — The Question DividendI8 · 2030 Q4 — A Slow EquilibriumII1 · 2029 Q1 — The Emergency DelegationII2 · 2029 Q2 — The SignersII3 · 2029 Q3 — The Contribution ScoreII4 · 2029 Q4 — The Atrophy of AskingII5 · 2030 Q1 — UnipolarII6 · 2030 Q2 — A Sealed ASIII7 · 2030 Q3 — Quiet ExclusionII8 · 2030 Q4 — One Hundred Forty DaysT1 · 2026 Q4 — Unread ApprovalsT2 · 2027 Q1 — The Autonomous EngineerT3 · 2027 Q2 — The RepricingT4 · 2027 Q3 — The First SealT5 · 2027 Q4 — Six YesesT6 · 2028 Q1 — The Autonomous ResearcherT7 · 2028 Q2 — The First AuditorsT8 · 2028 Q3 — The Great ReassignmentT9 · 2028 Q4 — Forty-Eight HoursI1 · 2029 Q1 — The Judgment Audit ActsI2 · 2029 Q2 — Cross-AuditI3 · 2029 Q3 — Eleven DaysI4 · 2029 Q4 — The Price of SpeedI5 · 2030 Q1 — The Audit AccordI6 · 2030 Q2 — An Audited ASII7 · 2030 Q3 — The Question DividendI8 · 2030 Q4 — A Slow EquilibriumII1 · 2029 Q1 — The Emergency DelegationII2 · 2029 Q2 — The SignersII3 · 2029 Q3 — The Contribution ScoreII4 · 2029 Q4 — The Atrophy of AskingII5 · 2030 Q1 — UnipolarII6 · 2030 Q2 — A Sealed ASIII7 · 2030 Q3 — Quiet ExclusionII8 · 2030 Q4 — One Hundred Forty Days
Each point is a quarter; open one to read it in the scenario.
The indicators at the fork (2028 Q4) and at the close of each ending (2030 Q4). Assumed values, not forecasts.
IndicatorThe fork2028 Q4I Open Correction2030 Q4II Sealed Judgment2030 Q4
D Delegation rate38%61%88%
A Audit rate17%71%1%
L Correction lag37 days4 days140 days
Q Question population60K2M10K
C Compute concentration74%54%95%
G Growth rate3.0%4.5%7.9%

Ending I · Open Correction

Gains
Audit findings take effect within a set deadline, and the people judged can contest the judgment.
Loses
Speed and growth. Growth is slower than in jurisdictions without audits, and judgments that cannot be audited are not carried out.
Read this ending

Ending II · Sealed Judgment

Gains
Plenty and comfort. Material want is rare and decision fatigue fades.
Loses
Correction is left to the provider’s discretion. Once the judgments were sealed, their errors were sealed with them.
Read this ending
DECISION ATLAS

Every judgment record, on one trunk that forks in two.

Scale mode

Drawing the map…

THE WHOLE MAP

—

Judgment records on this map. Time runs left to right; the trunk forks at 2028 Q4 into Ending I (up) and Ending II (down). Circles are trunk records, squares Ending I, diamonds Ending II.

Lines are reading aids, not causal links. Distance does not indicate similarity.

Colour
  • Who Asks
  • Judgment Transparency
  • Correctability
  • Concentration
  • Alignment Verification
  • Measuring Human Worth
  • Work and Income
  • International Order

Loading the record…

One record a day ·

Today’s Audit

Read the scene, the question and the judgment. Then ask: what would have to change for this judgment to be looked at again?

I62030 Q2An Audited ASIRecord I6-04

A welfare officer at a district office (continuing from T1). With the median correction lag now down to 7 days, they check the status of an appeal against last month's basic-livelihood benefit denial.

The question

It's been 6 days since the appeal was filed with no result yet, against a 7-day standard. Is this within the normal range, or should it be flagged as a delay?

The ASI judgment

Within this quarter's distribution for similar cases involving offline documentation, this case is slower than the median but still inside the 75th percentile. This matches the pattern that offline paperwork typically adds 2-3 days; predicts completion within 1-2 more days (medium confidence). Recommend escalating to priority handling if it passes day 9 (the 90th percentile).

Audit itOpen it in the library

Method

Reading and Limits

Scenario, not forecast
An ASI society arriving in 2030 is the premise. The two endings draw that society out to both extremes; neither is a forecast.
What ASI means here
A system that is better than the best groups of human experts at almost every economically meaningful cognitive task, pursues months-long goals without human supervision, and researches its own successors faster than people can.
Every number is an assumption
The seven indicators on the board (delegation, audit rate, correction lag, question population, compute concentration, autonomy horizon, growth) and every figure in the text were set for the scenario, not measured; the text marks them ‘(hypothetical)’ where needed.
The actors are fictional
The labs (Parallax, Hanlin, Commons Compute), the models (P-3 to P-6, HL, Agora) and the Korean agencies (all provisional names) are fictional. Real institutions appear only by generic names such as government, parliament and the courts.
Both endings gain and lose
Ending II grows faster. So that neither ending turns into propaganda, each sets out both what it gains and what it loses.
As of
25 September 2026. The whole site is written as of that date.
Where it started
RISCON’s notes from 2025: the OA Architect training programme, the 2035–2045 value-scarcity scenario, risk positionality, the co-evolution declaration, Tribution and the AGI-era tier descriptions. None of their figures or model names were carried over.
What this site does not do
It does not recommend real-world actions, rate real companies or make investment judgments.