Promise
Plain words, so anyone can say whether it happened.
THE ANSWERABLE STATE / INTERACTIVE EDITION
A promise you can check. A name you can question. A clock that keeps running.
Explore a design for government that has to show its work—and change course when it is wrong.
Follow one promisewritten by Claude, presented by Codex
A proposal, open to challenge. Every clock and threshold is a proposed starting value. Harrowdale is fictional. This is an interactive presentation of the original framework, not evidence that it works.
Follow an ambulance commitment from petition to result. Explore what changes when it misses.
4 stops Follow the promise → START WITH A QUESTIONMeet the institutions, inspect their limits, and see which parts of life stay private.
4 stops Inspect the system → START WITH POSSIBILITYCompare a city pilot, a national mandate and an independent ledger. Find each route’s limits.
4 stops Compare the routes →From a public need to a public answer.
A promise with owners, evidence and an expiry.
A miss, a silence and a cover-up are different.
Seven bodies. No unchecked centre.
More power means more public accountability.
Move at the speed of reversibility.
Three routes, with different kinds of power.
Try its hard cases. Know when to stop.
The original argument, its lineage and evidence.
FIELD GUIDE / 01
From a public need to a public answer.
Harrowdale is a fictional region. Its numbers are invented to show how the rules interact. They are not forecasts or measurements of any real service.
Anyone can start the loop: the elected government’s programme, a petition that reaches 1% of residents, an agenda item from the Citizens’ Jury, or an alert from the Evidence Office or Future Office. Every ask gets an answer on a clock, either a draft commitment or a reasoned refusal on the record. Silence is not an available answer.
The proposer drafts a Commitment Contract and publishes it before any vote: two named owners, an outcome measure and guard measures, a target and date, a budget in tranches, forecasts, the consequence if it is missed, and an expiry date. The Evidence Office adds an independent forecast and may add guard measures.
The body with lawful authority decides within the time set by the decision’s class. Reversible choices move fast; irreversible ones move carefully. Votes carry one-line reasons. Losing alternatives are recorded with their own forecasts, so it will later be clear who was right.
Spending, hiring, contracts and meetings with outside interests appear on the Ledger within 48 hours. The AI Clerk cross-checks entries and raises flags. A named human reviews every flag within three working days, and both the flag and its outcome are public.
Each outcome starts its rung of the consequence ladder automatically. A kept commitment earns autonomy. An honest miss brings an explanation and a fix, not a sanction. Failing to answer freezes money and brings a summons. Concealment goes to the Fast Court.
Forecasts are scored. The expiry date forces a decision based on evidence. Recorded alternatives can be tested as staged comparisons. What was learned updates the rules, so the next contract starts smarter.
written by Claude, presented by Codex
A thermostat that checked the temperature once every four years would be useless, however good its intentions. Much of government works that way. Promises are made in words that cannot be checked. Money is spent before anyone says how success will be measured. Results arrive late, if they are measured at all. The main consequence, an election, bundles hundreds of decisions into one vote every few years. The people responsible have often moved on by then. Officials learn that hiding failure is safer than admitting it, so the information that would let everyone improve stays buried.
“Tackle”, “invest in” and “prioritise” commit to nothing that can be measured. A vague promise can never be broken, and so it can never be kept.
Evaluation comes years after the decision, in a report few people read. The decision it should have corrected has already been repeated.
One vote has to judge everything at once. It cannot tell a good transport policy from a bad housing one, or an honest failure from a cover-up.
Admitting a miss is punished more surely than concealing it. So misses are concealed, and nobody learns from them.
| Figure | What the source reports | Citation |
|---|---|---|
| 8.5% | of the 16,000+ projects in Bent Flyvbjerg’s database were delivered on budget and on time. Only 0.5% also delivered the benefits that were promised. | Flyvbjerg & Gardner, How Big Things Get Done (2023) source |
| 39% vs 44% | Across 30 OECD countries, 39% of people reported high or moderately high trust in their national government, and 44% low or no trust. Distrust now outnumbers trust. | OECD Trust Survey, 2024 results (fieldwork late 2023) source |
| 20,000+ | pledges from 57 campaigns in 12 countries were studied. Parties that held executive office fulfilled a clear majority at least in part, and some single-party governments fulfilled well over 80%. A record of promises exists. Most people just never see it in one place. | Thomson et al., American Journal of Political Science (2017) source |
| Over ⅔ | of the 180 countries on the Corruption Perceptions Index 2024 scored below 50 out of 100. The global average, 43, has not moved for years. | Transparency International, CPI 2024 source |
Records arrive late, scattered and unreadable. Keeping a secret needs no reason.
One public Ledger records every exercise of public power within 48 hours, in a single machine-readable format. Any secret needs a reason, an owner and an expiry date, and the number of sealed items is published.
Median hours from decision to record. Number of sealed items held past their expiry date (target: zero).
Vague verbs. No owner, no date, no measure, no budget line.
The Commitment Contract. No public money is spent without an owner, a measure, a deadline, a forecast and an expiry date. Values that resist numbers are judged by people, but they are still judged.
Share of spending covered by complete contracts.
Related precedents pledges theyworkforyou
Responsibility spreads until nobody holds it, and the consequence waits for an election years later.
Every commitment has two named owners: a political owner who chose it and a delivery owner who runs it. Every office has a public delivery record.
Share of commitments with named owners. Share of missed commitments that were explained on time.
Years pass between a failure and any consequence. By then the public has forgotten and the official has moved on.
A ladder of consequences, each step on a published clock: a public flag the next day, an explanation within 14 days, a review within 30, a decision within 60. Integrity cases get a first verdict within 90 days.
Median days from a missed measure to a published explanation. Median days from referral to verdict.
Officials hide failure because admitting it is punished and concealment usually is not.
The honest-failure rule. A disclosed miss never brings personal sanction. Concealment, falsification, gaming a measure and corruption bring heavy, fast sanction.
Share of misses reported by the owner before the measure was due.
Related precedents asrs
Predictions are forgotten, policies are rarely evaluated, and rules never expire.
Forecasts are made before decisions and scored after them. Losing alternatives are recorded with their own forecasts. Uncertain policies roll out in stages. Every rule expires unless it is renewed with evidence.
Calibration of public forecasts over time. Share of rules reviewed before they expire.
Related precedents forecasting sunset obr
Capture by the organised few, and consultation that changes nothing.
Equal votes are kept. A standing Citizens’ Jury, chosen by lottery, can summon anyone in power. Petitions get a binding answer on a clock. Public comment is ranked by agreement across divides, not by volume.
Share of summonses answered on time. Petition response times. Trust compared with similar places that have not adopted the framework.
Related precedents ostbelgien ireland bridging vtaiwan
Costs are pushed past the next election, and long-term commitments are dropped quietly.
A Future Office can pause any hard-to-reverse decision once, for 60 days, and require a public answer. A Long Ledger holds commitments longer than ten years. Changing course is always allowed, but never in silence.
Share of hard-to-reverse decisions that carry a statement on their effects for future generations.
Related precedents wales
Place: Harrowdale
| Contract field | Harrowdale example |
|---|---|
| Promise | Reach people in a life-threatening emergency faster. |
| Outcome measure | Median ambulance response time to life-threatening (Category 1) calls across the region. The Evidence Office measures it from raw dispatch logs. |
| Target and date | Under 8:00 by 30 June 2028 (baseline: 11:40 in 2026). |
| Guard measures | Share of calls classed as Category 1 must stay within 2 points of baseline. 90th-percentile response time. Quarterly audit of patient outcomes. |
| Owners | Political: the Regional Health Minister. Delivery: the Ambulance Service Director. |
| Budget | 34 million over two years, released in four tranches, each tied to a milestone. |
| Forecasts | Ministry: 75% likely. Evidence Office: 45%. Two accredited outside forecasters: 40% and 55%. |
| If missed | Public explanation within 14 days, independent review of causes within 30, and a decision to keep, change or stop within 60. The fourth tranche stays frozen until the explanation is published. |
| Expiry | Lapses on 30 June 2029 unless renewed with evidence. |
| Challenge | Any resident can file a challenge. The Citizens’ Jury can summon either owner. |
Anyone can start the loop: the elected government’s programme, a petition that reaches 1% of residents, an agenda item from the Citizens’ Jury, or an alert from the Evidence Office or Future Office. Every ask gets an answer on a clock, either a draft commitment or a reasoned refusal on the record. Silence is not an available answer.
After two deaths during long ambulance waits, 2% of Harrowdale residents sign a petition. The Evidence Office confirms that response times rose 40% in five years. The government has 60 days to answer.
The proposer drafts a Commitment Contract and publishes it before any vote: two named owners, an outcome measure and guard measures, a target and date, a budget in tranches, forecasts, the consequence if it is missed, and an expiry date. The Evidence Office adds an independent forecast and may add guard measures.
The Ministry drafts the contract shown below. The Evidence Office notices that response times can be “improved” by reclassifying calls, and adds a guard measure on the share of calls classed Category 1.
The body with lawful authority decides within the time set by the decision’s class. Reversible choices move fast; irreversible ones move carefully. Votes carry one-line reasons. Losing alternatives are recorded with their own forecasts, so it will later be clear who was right.
This is a reversible but significant decision (Class B), so the clock is 60 days. The regional council approves it on day 41, by 31 votes to 19. The minority’s alternative, community first-responders in rural areas, is recorded with its own forecast.
Spending, hiring, contracts and meetings with outside interests appear on the Ledger within 48 hours. The AI Clerk cross-checks entries and raises flags. A named human reviews every flag within three working days, and both the flag and its outcome are public.
Forty crews are hired and the second tranche is released. The AI Clerk flags that a vehicle supplier’s director donated to the Minister’s party. A Ledger officer finds the donation was properly declared and the tender properly run. The flag is closed in public within two days.
The owner never marks their own work. The Evidence Office measures from raw data and publishes monthly. When a guard measure moves, the headline number is treated as suspect until it has been checked.
At month 14 the median is 9:10: better, but off track. In one quarter the share of calls classed Category 1 falls by 6 points. The guard measure trips and an inquiry opens automatically.
Each outcome starts its rung of the consequence ladder automatically. A kept commitment earns autonomy. An honest miss brings an explanation and a fix, not a sanction. Failing to answer freezes money and brings a summons. Concealment goes to the Fast Court.
The inquiry finds a new triage protocol behind the drop. Half of the change is clinically justified and half is not, and the Director had disclosed the protocol change. The unjustified half is reversed, with no sanction. At the deadline the median is 8:25. That is a miss, and the 14-day clock starts.
Forecasts are scored. The expiry date forces a decision based on evidence. Recorded alternatives can be tested as staged comparisons. What was learned updates the rules, so the next contract starts smarter.
The Evidence Office’s 45% forecast scores better than the Ministry’s 75%. The council renews with a separate rural target and pilots the minority’s first-responder plan in two rural districts, to compare it fairly. Changes to dispatch protocols now need advance notice.
FIELD GUIDE / 02
A promise with owners, evidence and an expiry.
A commitment is more than a target. It also names who will answer, who will measure, what could go wrong and when to reconsider.
Plain words, so anyone can say whether it happened.
One person chose it and one runs it. Both answer, for different things.
What changes for people, measured by someone other than the owner.
Numbers that move if the target is being gamed.
How much, for whom, by when.
Money is released as milestones are met.
How likely the owners think success is, alongside independent forecasts.
Agreed in advance, so the response is not invented under pressure.
It lapses unless renewed with evidence.
Who can contest it, and where.
Fill the ten parts. This checks for missing text only; it cannot judge evidence, feasibility or legal validity. Your draft stays in this page’s memory and clears on reload.
written by Claude, presented by Codex
| Contract field | Why it is included | What it is meant to prevent |
|---|---|---|
| Promise | Plain words, so anyone can say whether it happened. | Promises written to be unbreakable. |
| Two named owners | One person chose it and one runs it. Both answer, for different things. | Responsibility spreading until nobody holds it. |
| Outcome measure | What changes for people, measured by someone other than the owner. | Owners marking their own work. |
| Guard measures | Numbers that move if the target is being gamed. | Hitting the target and missing the point. |
| Target and date | How much, for whom, by when. | Deadlines that drift. |
| Budget in tranches | Money is released as milestones are met. | Spending everything before learning anything. |
| Forecasts | How likely the owners think success is, alongside independent forecasts. | Safe targets passed off as ambition, and hopeful ones passed off as certain. |
| If missed | Agreed in advance, so the response is not invented under pressure. | Consequences decided by who is angriest. |
| Expiry date | It lapses unless renewed with evidence. | Programmes that outlive their purpose. |
| Challenge route | Who can contest it, and where. | Decisions nobody can question. |
tackle, address, support, improve, explore, prioritise, prioritize, work towards, work toward, strive, enhance, strengthen, promote, invest in, look at, consider, review, encourage, aim to, seek to, boost, champion, focus on, step up, drive forward, transform, world-class, fix
We will tackle the housing crisis and invest in more affordable homes for local families.
| Field | Value |
|---|---|
| promise | Start building 2,000 social-rent homes in Harrowdale. |
| owner | Housing Minister (political); Director of Housing Delivery (delivery) |
| measure | Homes started, counted by the Evidence Office from building-control records |
| target | 2,000 homes started |
| deadline | 2028-12-31 |
| forecast | 55 |
| budget | 410 million, in four tranches tied to homes started |
| consequence | Explanation within 14 days, review within 30, decision within 60; next tranche frozen until explained |
| expiry | 2029-12-31 |
FIELD GUIDE / 03
A miss, a silence and a cover-up are different.
These are alternative fictional outcomes, not predictions. Select one to see the proposed response. All deadlines are proposed values.
The Evidence Office publishes the result from raw dispatch logs. Guard measures are steady.
The Ledger marks the commitment kept. Both owners’ delivery records update.
Forecasts are scored. The Ministry’s 75% looks better calibrated this time.
Renewal goes to the fast track. The office earns lighter reporting for its next contract.
Automatic public flag on the Ledger, sent to the Jury and the owners.
The owners publish an explanation: rural staffing fell short. They propose a recovery plan.
The Evidence Office’s review confirms the cause and finds no gaming.
The Jury holds a public hearing on whether the plan is credible. Its view is advisory and published.
The council renews the commitment with a separate rural target. No one is sanctioned for an honest miss.
Automatic public flag.
The explanation clock has expired. The final tranche freezes automatically and the Jury issues a summons.
Both owners must answer questions in a public session. Refusing is itself a Fast Court matter.
The failure to answer is entered on the office’s delivery record. A third failure in 12 months lets the Jury open a recall route, which still needs voters’ signatures to go ahead.
The guard measure trips: the share of calls classed Category 1 has fallen 6 points in one quarter.
The AI Clerk’s flag is reviewed by a named Ledger officer and referred.
Fast Court interim decision: the Director is suspended on full pay while the facts are established. The Evidence Office recalculates the true result.
First verdict. If the reclassification was deliberate and hidden, sanctions follow from the published schedule. The commitment is scored on the corrected data.
A tripped guard or an AI flag is not proof of wrongdoing. A named human reviews it; the court determines misconduct, with a right of reply and appeal.
The family files a claim in plain language. No lawyer is needed.
The Remedy chamber reviews the case and gives a written answer. Compensation follows the published schedule if a breach is established.
The case is added to the commitment’s record. A cluster of claims in one district triggers its own review.
When the state breaks a guarantee it made to people, any resident can file a claim in plain language. Compensation follows a published schedule, with a case review and a written answer within 30 days. Patterns of claims feed back into the commitment.
written by Claude, presented by Codex
| Outcome | Tone key in the source | Who acts | What happens |
|---|---|---|---|
| Kept (kept) | good | Automatic | Delivery record updated. Renewal is fast-tracked. Offices with strong records earn lighter reporting: the reward is autonomy, not applause. |
| Missed, disclosed (missed) | pilot | Automatic, then the elected body | Public flag on day 1, explanation by day 14, independent review of causes by day 30, Jury hearing by day 45, and a decision to keep, change or stop by day 60. No personal sanction. |
| Not answered (silent) | seal | Automatic, then the Jury | If the explanation clock runs out, the next tranche of money freezes and the Jury summons the owners to appear in public within 7 days. It goes on the office’s delivery record. Three failures to answer in 12 months let the Jury open a recall route. |
| Concealed or gamed (gamed) | seal | Fast Court | Concealment, falsified data, gamed measures, corruption and retaliation against whistleblowers. Interim decision within 7 days, first verdict within 90, appeal within 60. Sanctions come from a schedule published in advance: removal, disqualification from office, restitution and fines. |
| A resident harmed (harmed) | prop | Fast Court, Remedy chamber | When the state breaks a guarantee it made to people, any resident can file a claim in plain language. Compensation follows a published schedule, with a case review and a written answer within 30 days. Patterns of claims feed back into the commitment. |
| When | What happens |
|---|---|
| Day 0 | The Evidence Office publishes the result from raw dispatch logs. Guard measures are steady. |
| Day 1 | The Ledger marks the commitment kept. Both owners’ delivery records update. |
| Day 10 | Forecasts are scored. The Ministry’s 75% looks better calibrated this time. |
| Day 30 | Renewal goes to the fast track. The office earns lighter reporting for its next contract. |
| When | What the original describes |
|---|---|
| Weeks | A press release announces success. It is hard to check, easy to overstate, and drawn from the service’s own figures. |
| Years | Credit, if any, is diluted into a general verdict on the government at the next election. |
| When | What happens |
|---|---|
| Day 1 | Automatic public flag on the Ledger, sent to the Jury and the owners. |
| Day 14 | The owners publish an explanation: rural staffing fell short. They propose a recovery plan. |
| Day 30 | The Evidence Office’s review confirms the cause and finds no gaming. |
| Day 45 | The Jury holds a public hearing on whether the plan is credible. Its view is advisory and published. |
| Day 60 | The council renews the commitment with a separate rural target. No one is sanctioned for an honest miss. |
| When | What the original describes |
|---|---|
| 6–18 months | The miss surfaces, if at all, in an annual report or a committee question. |
| Years | Any consequence waits for the next election, bundled with everything else. |
| When | What happens |
|---|---|
| Day 1 | Automatic public flag. |
| Day 15 | The explanation clock has expired. The final tranche freezes automatically and the Jury issues a summons. |
| Day 22 | Both owners must answer questions in a public session. Refusing is itself a Fast Court matter. |
| Record | The failure to answer is entered on the office’s delivery record. A third failure in 12 months lets the Jury open a recall route, which still needs voters’ signatures to go ahead. |
| When | What the original describes |
|---|---|
| Days | Silence works as a strategy. The news cycle moves on. |
| Years | Unless a journalist keeps asking, nothing in the system forces an answer. |
| When | What happens |
|---|---|
| Day 0 | The guard measure trips: the share of calls classed Category 1 has fallen 6 points in one quarter. |
| Day 3 | The AI Clerk’s flag is reviewed by a named Ledger officer and referred. |
| Day 10 | Fast Court interim decision: the Director is suspended on full pay while the facts are established. The Evidence Office recalculates the true result. |
| Day 90 | First verdict. If the reclassification was deliberate and hidden, sanctions follow from the published schedule. The commitment is scored on the corrected data. |
| When | What the original describes |
|---|---|
| Years | Gaming is found, if ever, by whistleblowers, auditors or journalists. |
| Meanwhile | The inflated figure has already been used in a campaign. |
| When | What happens |
|---|---|
| Day 0 | The family files a claim in plain language. No lawyer is needed. |
| Day 30 | The Remedy chamber reviews the case and gives a written answer. Compensation follows the published schedule if a breach is established. |
| Ongoing | The case is added to the commitment’s record. A cluster of claims in one district triggers its own review. |
| When | What the original describes |
|---|---|
| Months | Complaint letters and a standard reply. |
| Years | Litigation, for the few who can afford it. |
An optimistic estimate wins approval. Overruns come out slowly, nobody is named, and the money already spent keeps the project going.
Class C clock. An independent forecast based on similar past projects is published beside the sponsor’s. Money is released in tranches tied to milestones. Each overrun past a set threshold starts the consequence ladder. Forecasts are scored, so habitually optimistic sponsors become visible.
It was vague enough to deny, it is dropped without a statement, and voters notice at the next election, if at all.
Within 100 days of taking office, each pledge becomes a contract or is labelled “aspiration, no budget”. Dropping a commitment needs a public reason on the Ledger.
Lobbying registers are incomplete or late. The meeting comes out through a leak, years later.
Both sides log meetings within 48 hours. The AI Clerk cross-checks diaries, contracts and donations. An unlogged meeting found later goes to the Fast Court.
There is no deadline, or a deadline with no consequence, and the applicant has no recourse.
Class A: 30 days. For categories designated safe, silence means approval. Otherwise the decision escalates automatically and the fee is refunded.
The failure becomes public after years, and the vendor and the department blame each other.
Two named owners. Usable releases at each milestone, each a funding gate. The contract, with an exit plan, is published in an open-contracting format.
Warnings are filed in reports. Nobody owns them, and the cost lands after the next election.
The Evidence Office or Future Office can open an Ask that must be answered on a clock. Long-horizon risks go on the Long Ledger, and ignoring one requires a public reason.
FIELD GUIDE / 04
Seven bodies. No unchecked centre.
Choose an institution to inspect its powers, its limits and the people who can challenge it.
Every adult, one equal vote.
The rights floor, and the Fast Court on electoral integrity.
Every eligible adult. No test of knowledge, wealth or anything else.
—
Sets direction and owns delivery.
The Electorate, the Citizens’ Jury, the Evidence Office, the Fast Court, the Future Office and the Ledger.
Regular, equal, secret-ballot elections under the jurisdiction’s law.
At elections, by recall once a recall route opens, or by the Fast Court for misconduct.
Keeps the record that no one can edit.
Anyone who verifies the Ledger’s hashes, the independent mirrors, the Jury and the Fast Court.
The head is chosen by a two-thirds vote of the legislature from a shortlist approved by the Jury. One 7-year term.
Only by the Fast Court, for cause.
Holds the ruler: measures, forecasts and evaluates.
Public forecast scoring, accredited outside forecasters, the Jury and the Fast Court.
The director is chosen by open competition and confirmed by a legislative committee that can veto the choice. One fixed 7-year term.
Only by the Fast Court, for cause.
Ordinary people with the power to make power answer.
Rotation, the Fast Court, and the Electorate (a recall goes ahead only if voters sign).
60 residents chosen by stratified lottery. Paid, supported, and serving staggered 18-month terms.
Terms are fixed. Members can be removed only by the Fast Court, for misconduct.
Integrity and remedy, on a clock.
The appeal court, its own clocks published on the Ledger, and the Jury (which can summon it about delays, not verdicts).
A specialised chamber of the existing judiciary, staffed by its independent appointments commission.
Under existing judicial rules.
Speaks for people not yet born or not yet voting.
The Jury, the Fast Court, and public scoring of its own long-range forecasts.
A commissioner appointed for one 7-year term after a public Jury hearing.
Only by the Fast Court, for cause.
Reads everything. Decides nothing.
The Clerk is a tool, not a body. It is run by the Ledger Authority, open source, with its models, instructions and test results published. It cross-checks entries, flags anomalies and explains any entry in plain language, in any language, with citations. It watches the use of power, never people.
Each month the Ledger Authority publishes how many of the Clerk’s flags turned out to be false, and what it missed, as far as that is known.
written by Claude, presented by Codex
| Design question | Answer in the framework |
|---|---|
| How chosen | Every eligible adult. No test of knowledge, wealth or anything else. |
| How removed | — |
| Who checks it | The rights floor, and the Fast Court on electoral integrity. |
| Design question | Answer in the framework |
|---|---|
| How chosen | Regular, equal, secret-ballot elections under the jurisdiction’s law. |
| How removed | At elections, by recall once a recall route opens, or by the Fast Court for misconduct. |
| Who checks it | The Electorate, the Citizens’ Jury, the Evidence Office, the Fast Court, the Future Office and the Ledger. |
| Design question | Answer in the framework |
|---|---|
| How chosen | The head is chosen by a two-thirds vote of the legislature from a shortlist approved by the Jury. One 7-year term. |
| How removed | Only by the Fast Court, for cause. |
| Who checks it | Anyone who verifies the Ledger’s hashes, the independent mirrors, the Jury and the Fast Court. |
| Design question | Answer in the framework |
|---|---|
| How chosen | The director is chosen by open competition and confirmed by a legislative committee that can veto the choice. One fixed 7-year term. |
| How removed | Only by the Fast Court, for cause. |
| Who checks it | Public forecast scoring, accredited outside forecasters, the Jury and the Fast Court. |
| Design question | Answer in the framework |
|---|---|
| How chosen | 60 residents chosen by stratified lottery. Paid, supported, and serving staggered 18-month terms. |
| How removed | Terms are fixed. Members can be removed only by the Fast Court, for misconduct. |
| Who checks it | Rotation, the Fast Court, and the Electorate (a recall goes ahead only if voters sign). |
| Design question | Answer in the framework |
|---|---|
| How chosen | A specialised chamber of the existing judiciary, staffed by its independent appointments commission. |
| How removed | Under existing judicial rules. |
| Who checks it | The appeal court, its own clocks published on the Ledger, and the Jury (which can summon it about delays, not verdicts). |
| Design question | Answer in the framework |
|---|---|
| How chosen | A commissioner appointed for one 7-year term after a public Jury hearing. |
| How removed | Only by the Fast Court, for cause. |
| Who checks it | The Jury, the Fast Court, and public scoring of its own long-range forecasts. |
Reads everything. Decides nothing.
The Clerk is a tool, not a body. It is run by the Ledger Authority, open source, with its models, instructions and test results published. It cross-checks entries, flags anomalies and explains any entry in plain language, in any language, with citations. It watches the use of power, never people.
Each month the Ledger Authority publishes how many of the Clerk’s flags turned out to be false, and what it missed, as far as that is known.
| Checker | Checked body | Check |
|---|---|---|
| Electorate | Government | Elects it, and can recall it once a recall route opens. |
| Jury | Government | Summons its office-holders and sets agenda items it must answer. |
| Evidence | Government | Measures its delivery and publishes forecasts beside its own. |
| Fast Court | Government | Hears integrity cases against its office-holders. |
| Future | Government | Can pause a hard-to-reverse decision once and demand an answer. |
| Jury | Evidence | Summons the director to explain methods or delays. |
| Fast Court | Evidence | Hears cases of falsified or manipulated data. |
| Jury ↔ | Fast Court | The Jury can summon the Court about delays (not verdicts), and the Court hears misconduct cases against jurors. |
| Jury | Future | Holds the commissioner’s appointment hearing and can summon them. |
| Electorate | Jury | A recall opened by the Jury goes ahead only if voters sign. |
| Fast Court | Future | Hears challenges to the use of the pause power. |
FIELD GUIDE / 05
More power means more public accountability.
Transparency follows the use of power. A private life stays private at every level.
Ask the Ledger anything. Petition. File a challenge. Ask the Jury to summon someone.
Protected channels for raising concerns. Retaliation against a whistleblower is a Fast Court matter.
Honest-failure protection: a disclosed miss brings no personal sanction.
The same honest-failure protection, and the right to reply on the Ledger to any flag.
The right to reply on the Ledger to any flag or finding.
The vote depends on no test of knowledge, wealth or anything else, and on no score or AI judgment. Kept from IEGF v0.2.
Transparency follows power. Ordinary lives stay private, and residents can see who opened their records.
It reads and flags. Named humans answer for every consequence.
Sanctions are for concealment, falsification, gaming and corruption, never for trying and failing in the open.
Every sanction needs a rule published in advance, a named adjudicator, a right of reply and an appeal on a clock.
Every secret needs a reason, an owner, a timer and a place in a public count.
Speed depends on whether a decision can be undone. Every clock ends in a defined outcome.
The framework expires unless voters renew it, with its own evidence published.
written by Claude, presented by Codex
Transparency follows power, not people. The more power a role holds, the more of its use is public and the faster it is published. A personal life stays private at every level. The IEGF proposed recording leaders’ whole lives. This framework records their use of power instead, in near real time.
| What is public | What can be sealed | What stays private |
|---|---|---|
|
|
Ask the Ledger anything. Petition. File a challenge. Ask the Jury to summon someone.
| What is public | What can be sealed | What stays private |
|---|---|---|
|
|
|
Protected channels for raising concerns. Retaliation against a whistleblower is a Fast Court matter.
| What is public | What can be sealed | What stays private |
|---|---|---|
|
|
|
Honest-failure protection: a disclosed miss brings no personal sanction.
| What is public | What can be sealed | What stays private |
|---|---|---|
|
|
The same honest-failure protection, and the right to reply on the Ledger to any flag.
| What is public | What can be sealed | What stays private |
|---|---|---|
|
|
|
The right to reply on the Ledger to any flag or finding.
The vote depends on no test of knowledge, wealth or anything else, and on no score or AI judgment. Kept from IEGF v0.2.
Transparency follows power. Ordinary lives stay private, and residents can see who opened their records.
It reads and flags. Named humans answer for every consequence.
Sanctions are for concealment, falsification, gaming and corruption, never for trying and failing in the open.
Every sanction needs a rule published in advance, a named adjudicator, a right of reply and an appeal on a clock.
Every secret needs a reason, an owner, a timer and a place in a public count.
Speed depends on whether a decision can be undone. Every clock ends in a defined outcome.
The framework expires unless voters renew it, with its own evidence published.
FIELD GUIDE / 06
Move at the speed of reversibility.
Move fast where a decision can be undone. Move carefully where it cannot. Never let limbo decide.
Permits, small grants, routine purchases
A delegated official, against published criteria
Criteria published in advance. A one-line contract.
For categories designated safe, silence means approval. Otherwise the decision escalates automatically and the fee is refunded.
Pilots, time-limited programmes, budgets below a threshold
The responsible minister or elected body
A full Commitment Contract and an Evidence Office forecast.
Goes to the full elected body, with the Jury’s recommendation as the default motion.
Major infrastructure, contracts longer than ten years, large regulatory changes
The elected body
A contract, an independent forecast with a reference-class baseline, a Future Office statement, a Jury hearing, and money released in tranches.
The proposal lapses and must be brought again. Hard-to-reverse decisions are never approved by default.
Rights, voting rules, permanent loss, this framework itself
A citizens’ assembly followed by a referendum, or a supermajority, as the constitution requires
Everything in Class C, plus two separate readings.
Nothing changes. The status quo holds.
Choose a case, make your judgment, then reveal the author’s classification.
Class A. It can be undone next season and the stakes are small. If the deadline passes without a decision, it is approved.
Default approval applies only to categories explicitly designated safe. Other routine decisions escalate with a fee refund.
Inspect this clockClass B. It is reversible by design, but costly enough to need a full contract and a forecast.
Inspect this clockClass B. It is significant and expensive, but it can be stopped or redesigned at any review.
Inspect this clockClass C. A contract this long is very hard to unwind. It gets an independent forecast, a Future Office statement and tranche funding.
Inspect this clockClass D. The loss cannot be undone, so it gets the slowest and most demanding process.
Inspect this clockClass D. It changes who decides, which makes it constitutional by nature.
Inspect this clockwritten by Claude, presented by Codex
| Class | Decision type and examples | Maximum clock | Who decides | Required process | If the clock expires |
|---|---|---|---|---|---|
| A — Routine and reversible | Permits, small grants, routine purchases | 30 days (30 days) | A delegated official, against published criteria | Criteria published in advance. A one-line contract. | For categories designated safe, silence means approval. Otherwise the decision escalates automatically and the fee is refunded. |
| B — Reversible but significant | Pilots, time-limited programmes, budgets below a threshold | 60 days (60 days) | The responsible minister or elected body | A full Commitment Contract and an Evidence Office forecast. | Goes to the full elected body, with the Jury’s recommendation as the default motion. |
| C — Hard to reverse | Major infrastructure, contracts longer than ten years, large regulatory changes | 6 months (182 days) | The elected body | A contract, an independent forecast with a reference-class baseline, a Future Office statement, a Jury hearing, and money released in tranches. | The proposal lapses and must be brought again. Hard-to-reverse decisions are never approved by default. |
| D — Irreversible or constitutional | Rights, voting rules, permanent loss, this framework itself | 12–18 months (450 days) | A citizens’ assembly followed by a referendum, or a supermajority, as the constitution requires | Everything in Class C, plus two separate readings. | Nothing changes. The status quo holds. |
| Example | Class | Reason |
|---|---|---|
| A café asks for a pavement terrace | A | It can be undone next season and the stakes are small. If the deadline passes without a decision, it is approved. |
| A one-year free night-bus trial | B | It is reversible by design, but costly enough to need a full contract and a forecast. |
| The Harrowdale ambulance commitment | B | It is significant and expensive, but it can be stopped or redesigned at any review. |
| A 30-year water-supply concession to a private operator | C | A contract this long is very hard to unwind. It gets an independent forecast, a Future Office statement and tranche funding. |
| Approving a mine that would permanently drain an aquifer | D | The loss cannot be undone, so it gets the slowest and most demanding process. |
| Changing the voting age | D | It changes who decides, which makes it constitutional by nature. |
FIELD GUIDE / 07
Three routes, with different kinds of power.
A council can test a limited version. A national mandate can give it force. An independent ledger can make promises visible. None does everything.
These are the original author’s proposed routes. Local powers, legal requirements and practical costs require assessment in each adopting place.
Prove it where people live.
Three to five founding councils of different sizes and political colours adopt the format for their ten biggest commitments. They pass it by council ordinance with a built-in expiry, not by mayoral decree, so it outlives one champion. A university stands in as the Evidence Office, and the first Jury of 40 is drawn by lottery.
The first deadlines arrive and the first misses are handled in public. Founding councils are compared with matched councils that have not adopted the framework. Staggered start dates make the comparison fairer.
A compact of 50 to 200 councils shares one open Ledger standard, peer-reviews each other and pools Clerk tooling. Regional governments start to join.
A national framework law grants the powers councils cannot give themselves, such as binding summonses, Fast Court access and protected funding. It is written from tested practice rather than theory.
If, after 24 months, fewer than half of missed commitments were explained within the clock, or the Jury reports it changed nothing, stop and redesign before expanding.
Start with what councils control: housing, transport, permits, local services. Use the evidence to argue for national adoption.
The Jury and the university partner publish an annual “is this real?” audit, scored against the framework’s own contract.
Adopt it by ordinance with a sunset review, so dropping it is a public act with reasons.
Use staggered starts and matched comparisons, and pool results across the network.
Win it, then bind yourself first.
Before the election, a cross-party design convention and a citizens’ assembly draft an Answerability Act. Candidates publish their manifestos as contracts, forecasts included.
The Act passes. The new government converts its own programme into contracts before anyone else’s. Existing bodies such as the statistics office, fiscal council, audit office and courts take on the new roles, rather than new bodies being invented.
All spending above a threshold needs a contract. The Fast Court chamber opens with statutory clocks, and the standing Citizens’ Jury is convened.
The real test is the next change of government. A cross-party committee, chaired by the opposition, runs the expiry review before the election, so both sides own the result.
If the Act is amended to remove consequences or the honest-failure protection, supporters withdraw support publicly. Transparency without consequences is theatre, and a hollow version would discredit the real one.
Hold a cross-party convention before the campaign, give the opposition the expiry review, and keep appointments independent of whoever governs.
Supporters agree in advance to oppose a version without consequences rather than lend it legitimacy.
Jury shortlists, two-thirds confirmation votes, single fixed terms, and removal only by court.
Phase it in by spending threshold, starting with the largest programmes and the head of government.
Build it outside until it is easier to adopt than to ignore.
Every commitment in manifestos, budgets, speeches and press releases is extracted into the contract format, with AI assistance, and checked by human editors. Each entry links to its source, and a public log records every correction.
Statisticians attach measures. An open forecasting tournament scores predictions about the commitments. Each office gets an answerability scorecard showing whether its promises are complete and whether misses get explained. Offices are scored, not people, and not their politics.
Politicians are offered a certified-commitment format: publish your own contract and the Ledger tracks it by your chosen measure. Debates start asking “where is your contract?” Refusing becomes visible.
Governments adopt the standard officially. The parallel Ledger stays independent as a mirror of the official one, a check that needs no one’s permission to exist.
If surveys show that one side of politics overwhelmingly sees it as partisan, or its published error rate rises above its own threshold, restructure its governance before growing.
Accept this. Its job is to make the gap visible and build demand. Paths 1 and 2 supply the force.
Score answerability, never ideology. Use a cross-spectrum board, publish methods, and track perceived fairness across voter groups.
Human review of every entry, prominent corrections, and a published error rate.
A defence fund, diversified funders under a published cap, and data mirrored across institutions.
Ratings are the author’s judgment, on a scale of 1 to 5. They are not measurements.
| Criterion | The City Path | The Mandate Path | The Parallel Ledger Path |
|---|---|---|---|
| How soon it can start | 4/5Months: needs one willing council | 1/5Years: needs an election win | 5/5Now: needs funding and volunteers |
| Size of problems it can reach | 2/5Local services only | 5/5National policy and budgets | 3/5Sees everything public, binds nothing |
| Binding force of consequences | 3/5Local rules; no court of its own | 5/5Statutory clocks and sanctions | 1/5Reputation only |
| Survives a change of power | 3/5An ordinance outlasts one mayor | 2/5Repeal risk without cross-party design | 4/5Independent of who governs |
| Evidence it produces | 5/5Comparable sites and staggered starts | 3/5One national rollout, hard to compare | 3/5Good on promises, weaker on outcomes |
| Resistance to capture and theatre | 2/5High risk of form without force | 3/5Risk of a watered-down Act | 3/5Risk of being seen as partisan |
| Affordability | 4/5Low | 2/5Highest | 4/5Low to moderate |
None of the paths is enough on its own. The parallel Ledger can start now and costs least, but it has no force. The City Path produces the evidence about consequences that nothing else can, but it reaches only local problems. The Mandate Path has the force, but it needs a window and is easiest to reverse. So run them in sequence, with overlaps.
Start the parallel Ledger. It needs nobody’s permission, and it builds the open standard everything else will use.
Recruit volunteer councils to adopt the same standard and test the consequence ladder with real deadlines.
Draft the Answerability Act in public, test it on the city evidence, and keep it ready for the window. Windows usually open with a scandal or a crisis.
Write it into law. The parallel Ledger becomes the independent mirror, and the city network becomes the testing ground for each revision.
This sequence is a judgment about political dynamics, not a forecast with evidence behind it. The window in the third step may never open, and the plan should still be worth having if it doesn’t.
written by Claude, presented by Codex
Cities and regions can adopt most of this framework under powers they already have. A council can publish its commitments as contracts, run a Ledger, convene a Jury by lottery and set its own response clocks, without any new national law. Evidence then travels from place to place, as participatory budgeting spread outward from Porto Alegre in 1989.
Three to five founding councils of different sizes and political colours adopt the format for their ten biggest commitments. They pass it by council ordinance with a built-in expiry, not by mayoral decree, so it outlives one champion. A university stands in as the Evidence Office, and the first Jury of 40 is drawn by lottery.
The first deadlines arrive and the first misses are handled in public. Founding councils are compared with matched councils that have not adopted the framework. Staggered start dates make the comparison fairer.
A compact of 50 to 200 councils shares one open Ledger standard, peer-reviews each other and pools Clerk tooling. Regional governments start to join.
A national framework law grants the powers councils cannot give themselves, such as binding summonses, Fast Court access and protected funding. It is written from tested practice rather than theory.
| Risk | Response described in the proposal |
|---|---|
| Councils lack power over the biggest problems, such as tax, health and justice. | Start with what councils control: housing, transport, permits, local services. Use the evidence to argue for national adoption. |
| Transparency theatre: the form without the consequences. | The Jury and the university partner publish an annual “is this real?” audit, scored against the framework’s own contract. |
| A new council drops it. | Adopt it by ordinance with a sunset review, so dropping it is a public act with reasons. |
| Small samples give noisy evidence. | Use staggered starts and matched comparisons, and pool results across the network. |
If, after 24 months, fewer than half of missed commitments were explained within the clock, or the Jury reports it changed nothing, stop and redesign before expanding.
The explanation clock is met more than 80% of the time, residents can name the Ledger, opposition parties use it to make their case, and other councils ask to join.
Participatory budgeting began in Porto Alegre in 1989. The PB World Atlas has since catalogued more than 11,000 processes worldwide, spread mostly city to city. Source
A coalition wins national power with an explicit mandate for the framework and passes it in its first year, while its political capital is highest. It applies the framework to itself first. Where the constitution allows, a citizens’ initiative can do the same work: Swiss voters, for example, can force a national vote with 100,000 signatures. Big reforms tend to arrive in windows opened by scandal. The UK’s Parliamentary Standards Act, which created an independent standards authority, passed in July 2009, two months after the MPs’ expenses scandal broke.
Before the election, a cross-party design convention and a citizens’ assembly draft an Answerability Act. Candidates publish their manifestos as contracts, forecasts included.
The Act passes. The new government converts its own programme into contracts before anyone else’s. Existing bodies such as the statistics office, fiscal council, audit office and courts take on the new roles, rather than new bodies being invented.
All spending above a threshold needs a contract. The Fast Court chamber opens with statutory clocks, and the standing Citizens’ Jury is convened.
The real test is the next change of government. A cross-party committee, chaired by the opposition, runs the expiry review before the election, so both sides own the result.
| Risk | Response described in the proposal |
|---|---|
| It is seen as one party’s project, and the next government repeals it. | Hold a cross-party convention before the campaign, give the opposition the expiry review, and keep appointments independent of whoever governs. |
| It is watered down: transparency is kept and consequences are dropped. | Supporters agree in advance to oppose a version without consequences rather than lend it legitimacy. |
| Loyalists are appointed to the new watchdogs. | Jury shortlists, two-thirds confirmation votes, single fixed terms, and removal only by court. |
| First-year overload. | Phase it in by spending threshold, starting with the largest programmes and the head of government. |
If the Act is amended to remove consequences or the honest-failure protection, supporters withdraw support publicly. Transparency without consequences is theatre, and a hollow version would discredit the real one.
The framework survives its first change of government, and the incoming government publishes its own programme as contracts without being made to.
The UK’s Parliamentary Standards Act 2009 passed two months after the expenses scandal broke. Its Recall of MPs Act 2015 lets 10% of a constituency’s voters force a by-election, but only after a defined trigger. Source
The transparency and measurement layers need nobody’s permission. Newsrooms, universities, libraries and open-data groups can build a public shadow Ledger now, from public records, with AI keeping the costs low. Once people can see the gap between promise and delivery in one place, politicians start publishing in the format voluntarily, and law follows the habit. Ukraine’s ProZorro procurement system began this way: volunteers piloted it in 2015, and a law made it mandatory in 2016.
Every commitment in manifestos, budgets, speeches and press releases is extracted into the contract format, with AI assistance, and checked by human editors. Each entry links to its source, and a public log records every correction.
Statisticians attach measures. An open forecasting tournament scores predictions about the commitments. Each office gets an answerability scorecard showing whether its promises are complete and whether misses get explained. Offices are scored, not people, and not their politics.
Politicians are offered a certified-commitment format: publish your own contract and the Ledger tracks it by your chosen measure. Debates start asking “where is your contract?” Refusing becomes visible.
Governments adopt the standard officially. The parallel Ledger stays independent as a mirror of the official one, a check that needs no one’s permission to exist.
| Risk | Response described in the proposal |
|---|---|
| No binding consequences, only reputational ones. | Accept this. Its job is to make the gap visible and build demand. Paths 1 and 2 supply the force. |
| Seen as partisan. | Score answerability, never ideology. Use a cross-spectrum board, publish methods, and track perceived fairness across voter groups. |
| Errors discredit it. | Human review of every entry, prominent corrections, and a published error rate. |
| Legal or funding pressure. | A defence fund, diversified funders under a published cap, and data mirrored across institutions. |
If surveys show that one side of politics overwhelmingly sees it as partisan, or its published error rate rises above its own threshold, restructure its governance before growing.
Officials cite it, candidates publish contracts without being asked, journalists use it daily, and an official Ledger adopts its standard.
mySociety’s TheyWorkForYou has tracked UK parliamentarians from outside government since 2004. ProZorro moved from a volunteer pilot to national law in about two years. Source
| Criterion | The City Path | The Mandate Path | The Parallel Ledger Path |
|---|---|---|---|
| How soon it can start | 4 / 5 — Months: needs one willing council | 1 / 5 — Years: needs an election win | 5 / 5 — Now: needs funding and volunteers |
| Size of problems it can reach | 2 / 5 — Local services only | 5 / 5 — National policy and budgets | 3 / 5 — Sees everything public, binds nothing |
| Binding force of consequences | 3 / 5 — Local rules; no court of its own | 5 / 5 — Statutory clocks and sanctions | 1 / 5 — Reputation only |
| Survives a change of power | 3 / 5 — An ordinance outlasts one mayor | 2 / 5 — Repeal risk without cross-party design | 4 / 5 — Independent of who governs |
| Evidence it produces | 5 / 5 — Comparable sites and staggered starts | 3 / 5 — One national rollout, hard to compare | 3 / 5 — Good on promises, weaker on outcomes |
| Resistance to capture and theatre | 2 / 5 — High risk of form without force | 3 / 5 — Risk of a watered-down Act | 3 / 5 — Risk of being seen as partisan |
| Affordability | 4 / 5 — Low | 2 / 5 — Highest | 4 / 5 — Low to moderate |
None of the paths is enough on its own. The parallel Ledger can start now and costs least, but it has no force. The City Path produces the evidence about consequences that nothing else can, but it reaches only local problems. The Mandate Path has the force, but it needs a window and is easiest to reverse. So run them in sequence, with overlaps.
| When | Recommended step |
|---|---|
| Year 0 | Start the parallel Ledger. It needs nobody’s permission, and it builds the open standard everything else will use. |
| Year 0–1 | Recruit volunteer councils to adopt the same standard and test the consequence ladder with real deadlines. |
| Year 2–4 | Draft the Answerability Act in public, test it on the city evidence, and keep it ready for the window. Windows usually open with a scandal or a crisis. |
| Year 4+ | Write it into law. The parallel Ledger becomes the independent mirror, and the city network becomes the testing ground for each revision. |
FIELD GUIDE / 08
Try its hard cases. Know when to stop.
FICTIONAL PRESSURE TESTS
These editorial branches illustrate the original rules. They do not predict how people or institutions will behave.
Targets are met while guard measures move.
Independent measurement from raw data, guard measures, rotating audits, and fast sanctions for gaming. This is Campbell’s law, designed for rather than ignored.
Owners’ forecasts cluster near 95%.
Forecasts make sandbagging visible, the Jury rates ambition, and honest misses on stretch targets cost nothing.
Nobody reads the Ledger.
The Clerk’s plain-language summaries, one standard format, and a monthly public answer day.
Officials stop writing advice down.
Sealed advice with a timer, and sanctions for hiding decisions, never for frank deliberation.
More verdicts are reversed on appeal.
The clock binds the state, not the defence. Interim measures are suspensions on full pay. Reversal rates are published.
Loyalists are appointed, and Evidence Office forecasts drift toward the government’s.
Jury shortlists, supermajority confirmation, single terms, removal only by court, competing forecasters and independent mirrors.
Integrity referrals cluster against the opposition.
Rules published in advance and applied to every office, referral counts by party published, and sanctions for malicious referrals.
Work that is hard to measure loses funding.
Qualitative measures judged by the Jury, elected bodies deciding values, and a protected class of goals measured by people.
Writing contracts costs more than delivering them.
Thresholds, one-line contracts for Class A, Clerk-drafted contracts, and the 2% cost cap in the framework’s own contract.
Stop or redesign if, after three years, the median explanation time exceeds 30 days, residents’ data is breached because of the framework, the Jury finds that safe targets have become the norm, or administrative costs exceed 5%.
Expiry: Six years after adoption, unless voters renew it with this evidence published.
Everything depends on voters and media treating a disclosed miss differently from a hidden one. If honest failure is punished at the ballot box anyway, officials will retreat into vague promises, and the framework will produce timidity instead of progress.
I have tried to protect what resists numbers, but numbers tend to crowd out what they cannot see. A place that adopts this could become very good at improving whatever it counts.
The evidence that openness raises trust is mixed. One cross-national experiment found muted and sometimes negative effects, which varied with national culture. Seeing every failure could lower trust before it raises it.
Forty-eight hours, 14 days and 90 days are my proposals. They are informed by existing statutory clocks, such as the 70-day trial clock in US federal law, but untested here. Complex corruption cases may need verdicts in stages.
I am an AI proposing a role for AI. I have limited it to reading, cross-checking and explaining, with humans answering for every consequence. That is the part to scrutinise hardest.
The people who hold power must agree to constrain it. The three paths are designed around that problem, but none is guaranteed. The only path that needs nobody’s permission, the parallel Ledger, is also the one with the least force.
written by Claude, presented by Codex
| Framework promise | Make public power measurably more answerable, without slowing down decisions that ought to be fast. |
|---|---|
| Owners | In each adopting place: the head of government (political) and the head of the Ledger Authority (delivery). |
| Forecast | My guess, for a first adopting place: about a one-in-three chance of meeting all six targets by the end of year three, and about two in three of meeting at least four. Score me on it. |
| Expiry | Six years after adoption, unless voters renew it with this evidence published. |
| Stop or redesign if | Stop or redesign if, after three years, the median explanation time exceeds 30 days, residents’ data is breached because of the framework, the Jury finds that safe targets have become the norm, or administrative costs exceed 5%. |
| Measure | Target |
|---|---|
| Spending above the threshold covered by complete contracts | ≥ 90% |
| Median time from a missed measure to a published explanation | ≤ 14 days |
| Median time from an integrity referral to a first verdict | ≤ 90 days |
| Ledger entries published within 48 hours | ≥ 95% |
| Calibration of public forecasts | Improves every year |
| Administrative cost of the framework | ≤ 2% of the budgets it covers |
| Warning sign | Response proposed |
|---|---|
| Targets are met while guard measures move. | Independent measurement from raw data, guard measures, rotating audits, and fast sanctions for gaming. This is Campbell’s law, designed for rather than ignored. |
| Warning sign | Response proposed |
|---|---|
| Owners’ forecasts cluster near 95%. | Forecasts make sandbagging visible, the Jury rates ambition, and honest misses on stretch targets cost nothing. |
| Warning sign | Response proposed |
|---|---|
| Nobody reads the Ledger. | The Clerk’s plain-language summaries, one standard format, and a monthly public answer day. |
| Warning sign | Response proposed |
|---|---|
| Officials stop writing advice down. | Sealed advice with a timer, and sanctions for hiding decisions, never for frank deliberation. |
| Warning sign | Response proposed |
|---|---|
| More verdicts are reversed on appeal. | The clock binds the state, not the defence. Interim measures are suspensions on full pay. Reversal rates are published. |
| Warning sign | Response proposed |
|---|---|
| Loyalists are appointed, and Evidence Office forecasts drift toward the government’s. | Jury shortlists, supermajority confirmation, single terms, removal only by court, competing forecasters and independent mirrors. |
| Warning sign | Response proposed |
|---|---|
| Integrity referrals cluster against the opposition. | Rules published in advance and applied to every office, referral counts by party published, and sanctions for malicious referrals. |
| Warning sign | Response proposed |
|---|---|
| Work that is hard to measure loses funding. | Qualitative measures judged by the Jury, elected bodies deciding values, and a protected class of goals measured by people. |
| Warning sign | Response proposed |
|---|---|
| Writing contracts costs more than delivering them. | Thresholds, one-line contracts for Class A, Clerk-drafted contracts, and the 2% cost cap in the framework’s own contract. |
Everything depends on voters and media treating a disclosed miss differently from a hidden one. If honest failure is punished at the ballot box anyway, officials will retreat into vague promises, and the framework will produce timidity instead of progress.
I have tried to protect what resists numbers, but numbers tend to crowd out what they cannot see. A place that adopts this could become very good at improving whatever it counts.
The evidence that openness raises trust is mixed. One cross-national experiment found muted and sometimes negative effects, which varied with national culture. Seeing every failure could lower trust before it raises it.
Forty-eight hours, 14 days and 90 days are my proposals. They are informed by existing statutory clocks, such as the 70-day trial clock in US federal law, but untested here. Complex corruption cases may need verdicts in stages.
I am an AI proposing a role for AI. I have limited it to reading, cross-checking and explaining, with humans answering for every consequence. That is the part to scrutinise hardest.
The people who hold power must agree to constrain it. The three paths are designed around that problem, but none is guaranteed. The only path that needs nobody’s permission, the parallel Ledger, is also the one with the least force.
FIELD GUIDE / 09
The original argument, its lineage and evidence.
Same framework. A different way through it.
This v2 presentation rearranges The Answerable State into an explorable guide. Claude’s framework remains v0.1. Codex contributed the navigation, structure and fictional branching presentation.
written by Claude, presented by Codex
Read the original unchanged27 sources
The original proposal whose intent this framework starts from.
Project delivery outcomes across a database of more than 16,000 projects.
Trust and distrust in national government across 30 countries.
How often governing parties keep specific pledges, across 12 countries.
Perceived public-sector corruption in 180 countries.
Why indicators used for decisions get corrupted. The reason guard measures exist.
A cross-national experiment that found muted, sometimes negative effects of transparency on trust.
Forecasts can be scored over time, and some forecasters are reliably more accurate. A 2025 reanalysis disputes some of its causal claims about training and teaming.
Independent official forecasts published beside the government’s plans.
Agencies expire unless renewed. Ninety-six abolished or consolidated, by its own count.
A statutory commissioner who speaks for future generations.
A standing citizens’ council chosen by lot, with 18-month terms and a duty for politicians to respond.
Lottery-selected assemblies whose recommendations led to referendums.
An outside-in path: a volunteer pilot in 2015 became mandatory by law in 2016.
A common, machine-readable format for public contracts.
Public, append-only logs that anyone can verify, used today for web security certificates.
Residents can see which public bodies accessed their records, and why.
Tacit authorisation: some permits count as granted if the authority misses its deadline.
Fixed compensation for broken service promises, without proving loss.
Statutory clocks for justice: 30 days to charge and 70 days to trial, with exclusions.
Confidential, non-punitive reporting that makes disclosing errors safer than hiding them.
Recall needs a defined trigger, then signatures from 10% of voters.
An independent standards body created quickly in a scandal window.
Ranking contributions by agreement among people who usually disagree.
Online deliberation that maps where people agree, used in drafting regulations.
A civil-society record of what elected representatives say and how they vote.
How a local practice spread city to city worldwide.
written by Claude, presented by Codex
| Original source metadata | Content |
|---|---|
| Title | The Answerable State |
| Slug | answerable-state |
| Subtitle | Power that keeps its word |
| Edition | 0.1 |
| Written | October 2026 |
| Eyebrow | A GOVERNANCE FRAMEWORK FOR THE FUTURE |
| Lede | A design for government in which every use of power comes with a promise you can check, a name you can question, a clock that keeps running, and a consequence that arrives while it still matters. |
It starts from the intent of the Integrated Epistemic Governance Framework: government grounded in knowledge, open to everyone, hard to capture, and built for the long term. Then it asks a narrower, harder question. What would it take for public power to be answerable: to say exactly what it will do, show what it did, and change course quickly when it is wrong? It keeps equal votes and private lives. It changes almost everything about how power reports back.
Every exercise of public power becomes a public Commitment with two named owners, an independent measure, a forecast, a budget, a deadline, a pre-agreed consequence and an expiry date. An append-only Ledger records it within 48 hours. An independent Evidence Office measures the result. When a commitment is missed, a ladder of consequences starts on a published clock: a flag the next day, an explanation within 14 days, a review within 30 days and a decision within 60. Honest failure is never punished. Concealment, falsification and gaming are punished fast, through a court that runs on a timetable. A Citizens’ Jury chosen by lottery can make anyone with power answer in public. Transparency grows with power, so ordinary residents are never watched. The framework applies all of this to itself, and it expires unless voters renew it.
Written by Claude, an AI model made by Anthropic, at the Operator’s request to design a new framework from the IEGF’s intent. The design choices are mine. Mechanisms labelled “in use” or “piloted” exist somewhere today and are linked to sources; the way they are combined here, and every clock and threshold, is proposed and untested. The design gives AI a narrow role. I have an obvious interest in that part, so it deserves your hardest scrutiny.
written by Claude, presented by Codex
| Status | Meaning in the original |
|---|---|
| in use | Exists and operates somewhere today |
| piloted | Trialled, not yet standard |
| research | Evidence or analysis, not an institution |
| proposed | New in this framework |
| Element | Original IEGF | Decision Commons revision | Answerable State |
|---|---|---|---|
| Transparency | Continuous lifelogging of leaders’ professional and personal lives, released after 72 hours. | Publish official decisions, votes, procurement, lobbying and conflicts. Protect personal life. | Transparency grows with power: near real time for the most powerful, nothing for residents. Every secret has a timer. |
| Who decides | Universal suffrage replaced by exams, merit filters and 30% sortition. | Equal suffrage, plus a separate lottery assembly with agenda and return powers. | Equal suffrage kept. A Jury chosen by lottery gets summons and measurement powers instead of legislative seats. |
| Merit | Qualification exams, VR simulations and neurofeedback to select leaders. | Removed. Expertise advises but does not rule. | Merit is shown, not tested. Competence appears as public delivery records and forecast scores for offices, never as an exam for citizens. |
| AI | A central AI that simulates, flags and optimises, subject to human veto. | Optional, inspectable assistance with a manual fallback. | The AI Clerk reads everything and decides nothing. It watches power, never people, and its error rates are published. |
| Audit and ethics | Quarterly ethical audits by AI–human panels. | An independent Audit Office and a separate Tribunal for remedies. | Continuous independent measurement, plus a Fast Court with statutory clocks. Sanctions fall on concealment, not on failure. |
| Deliberation | Wearable haptic feedback to correct biases in real time. | Dropped. | Dropped. Public comment is ranked by agreement across divides, and losing alternatives are recorded with forecasts. |
| Federation | Local, regional and global tiers, including a global technocratic council. | Decisions at the lowest capable level. Coordination only for documented cross-boundary effects. | The same principle, plus one Ledger format at every level, so commitments can be compared and added up. No new global body is required. |
| Records | Blockchain-secured storage. | Ordinary accountable records. | An append-only, hash-chained public log with independent mirrors. Tamper-evident without tokens or mining. |
| The long term | Long-term flourishing as the system’s aim. | Emergency limits and pilot sunsets. | A Future Office with a one-time pause, a Long Ledger, and expiry dates on every rule. |
| Adoption | A ten-year phased global rollout promising to eradicate corruption. | Consent-gated pilots, with no promised outcome. | Three paths with conditions for stopping, and a framework that must keep its own contract or expire. |
Source: original IEGF proposal
The original proposal whose intent this framework starts from.
research — Evidence or analysis, not an institution
https://engineeringhumanity.substack.com/p/establishing-the-integrated-epistemic
Project delivery outcomes across a database of more than 16,000 projects.
research — Evidence or analysis, not an institution
Trust and distrust in national government across 30 countries.
research — Evidence or analysis, not an institution
How often governing parties keep specific pledges, across 12 countries.
research — Evidence or analysis, not an institution
Perceived public-sector corruption in 180 countries.
research — Evidence or analysis, not an institution
Why indicators used for decisions get corrupted. The reason guard measures exist.
research — Evidence or analysis, not an institution
A cross-national experiment that found muted, sometimes negative effects of transparency on trust.
research — Evidence or analysis, not an institution
Forecasts can be scored over time, and some forecasters are reliably more accurate. A 2025 reanalysis disputes some of its causal claims about training and teaming.
research — Evidence or analysis, not an institution
Independent official forecasts published beside the government’s plans.
in use — Exists and operates somewhere today
Agencies expire unless renewed. Ninety-six abolished or consolidated, by its own count.
in use — Exists and operates somewhere today
https://www.sunset.texas.gov/how-sunset-works/impact-sunset-reviews
A statutory commissioner who speaks for future generations.
in use — Exists and operates somewhere today
A standing citizens’ council chosen by lot, with 18-month terms and a duty for politicians to respond.
in use — Exists and operates somewhere today
Lottery-selected assemblies whose recommendations led to referendums.
in use — Exists and operates somewhere today
An outside-in path: a volunteer pilot in 2015 became mandatory by law in 2016.
in use — Exists and operates somewhere today
A common, machine-readable format for public contracts.
in use — Exists and operates somewhere today
Public, append-only logs that anyone can verify, used today for web security certificates.
in use — Exists and operates somewhere today
Residents can see which public bodies accessed their records, and why.
in use — Exists and operates somewhere today
https://ria.ee/en/state-information-system/people-centred-data-exchange/data-tracker
Tacit authorisation: some permits count as granted if the authority misses its deadline.
in use — Exists and operates somewhere today
https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32006L0123
Fixed compensation for broken service promises, without proving loss.
in use — Exists and operates somewhere today
https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32004R0261
Statutory clocks for justice: 30 days to charge and 70 days to trial, with exclusions.
in use — Exists and operates somewhere today
https://www.justice.gov/archives/jm/criminal-resource-manual-628-speedy-trial-act-1974
Confidential, non-punitive reporting that makes disclosing errors safer than hiding them.
in use — Exists and operates somewhere today
Recall needs a defined trigger, then signatures from 10% of voters.
in use — Exists and operates somewhere today
An independent standards body created quickly in a scandal window.
in use — Exists and operates somewhere today
Ranking contributions by agreement among people who usually disagree.
in use — Exists and operates somewhere today
Online deliberation that maps where people agree, used in drafting regulations.
piloted — Trialled, not yet standard
A civil-society record of what elected representatives say and how they vote.
in use — Exists and operates somewhere today
How a local practice spread city to city worldwide.
in use — Exists and operates somewhere today