**On Legibility, Sunset Clauses, and the Perimeter That Outlives Its Covenant**
*A systems history, continuous with [Smart Prisons, Distributed Custody, and the Federal Interagency Reentry Council](https://bryantmcgill.blogspot.com/2026/07/distributed-custody.html), [Climate & Meritocracy: How Public Weather Data Became Private Risk Scores](https://bryantmcgill.blogspot.com/2025/11/climate-meritocracy.html), and [Cognitive-Cyber Warfare: Measures and Countermeasures](https://bryantmcgill.blogspot.com/2026/06/cognitive-cyber-warfare.html)*
This essay is a systems history of **the perimeter that survives the prison, the program, the administration, and even the legal instrument that created it**. It follows the evolution of correction from a bounded place into a distributed architecture of observation, identity resolution, scoring, intervention, differential access, and recursively measured response—an architecture that increasingly governs people through housing, employment, treatment, credit, mobility, information, and administrative status rather than through walls alone. Its central finding is not that institutions secretly coordinated to construct an invisible prison, but that **no leap is required when the topology is sufficient**: once multiple systems share the same operational grammar, their outputs can converge without their participants ever perceiving the whole.
The recurring law is simple: **the covenant is temporary, the sensorium is retained, and the ledger is durable**. Cambridge Analytica dissolved after consent was violated and publicly revoked, but the psychometric science, identity-resolution methods, inference systems, and targeting stack survived. Executive Order 13826 created a federal interagency reentry council with its own termination clause; under Section 6, the council expired on **7 March 2021**, while PATTERN continued scoring the federal prison population and the First Step Act’s risk-and-needs machinery remained operational. The Chan Zuckerberg Initiative announced the divestment of its criminal-justice portfolio in **January 2021**, transferring the visible justice covenant elsewhere, and in **March 2023** announced a new biological measurement program in Chicago built around thousands of sensors and sampling probes embedded in tissue. These events were not coordinated. Their significance lies in the recurrent institutional asymmetry they expose: **missions, councils, philanthropies, consent regimes, and reform programs decay faster than the measurement infrastructures, data standards, model weights, registries, and scoring systems created beneath them**.
This differential decay is the principal mechanism of invisibility examined here. Institutional bodies repeatedly change names, mandates, custodians, and documentary locations, leaving behind dead search strings and fragmented archives, while the functional architecture continues operating across successor programs. The Federal Interagency Reentry Council becomes the Federal Interagency Council on Crime Prevention and Improving Reentry, then the Reentry Coordination Council, then the Alternatives and Reentry Committee. The charter disappears; the function persists. The legal wrapper sunsets; the scoring engine continues. The institution stops helping before it stops measuring.
The essay reconstructs the longer genealogy of that architecture: the electronic tether invented in the 1960s as a **reward channel** for behavioral rehabilitation, inverted in the 1980s into an electronic handcuff, and later reintroduced as a humane alternative to incarceration; the transformation of federal probation into a results-driven control loop of observation, classification, dosage, intervention, and reassessment; the replacement of fixed temporal punishment with modeled thresholds of behavioral change; the expansion of custody from a physical location into an administrative status; and the construction of a commercial and governmental circulatory system through which claims about persons become portable long after the persons themselves have been released.
The article also identifies the precise point at which inference becomes coercion. Observation alone is not imprisonment. Neither is classification, modeling, or ranking. The perimeter appears when a score is coupled to **eligibility, pricing, restriction, intervention, or the graduated provisioning of affordance**. At that point, the model does not merely describe a future; it helps produce one. Adverse classification narrows opportunity, increases observation, reduces adaptive capacity, generates adverse outcomes, and then interprets those outcomes as confirmation of the original classification. Favorable classification expands opportunity and produces the opposite loop. The system appears meritocratic because outcomes correlate with prior rankings, while concealing that the ranking itself altered the conditions under which those outcomes became possible.
This is therefore not primarily an argument about surveillance, technology, prisons, philanthropy, finance, or artificial intelligence considered separately. It is an account of a common **risk-governance operating system** that ranks persons, properties, territories, institutions, and capital flows according to modeled future performance and then modulates how quickly, how freely, and under what conditions they may move. *Climate & Meritocracy* named the ledger. **This essay names the perimeter—and explains how it survives the expiration of its own covenant.**
## PART ONE — THE ADJACENCIES
### I. The First Adjacency
In January 2020, I was included by the [Global Justice Resource Center](https://globaljusticerc.org/)—with [a dedicated profile at globaljusticerc.org/bryant-mcgill](https://globaljusticerc.org/bryant-mcgill/)—inside a resource network concerned with justice reform, corrections, prison rehabilitation, technology, and international institutional change. That placement becomes more interesting when viewed through the people surrounding it: Executive Director [John L. Gannon, Ph.D.](https://www.linkedin.com/in/john-l-gannon-ph-d-897b8b17/) brought an unusual synthesis of philosophy, psychology, forensic clinical practice, correctional administration, prison-program development, and international forensic psychology, while President Uju Agomoh brought experience spanning UNODC consultation, Nigerian human-rights oversight, police and prison reform, and grassroots activism. Among the Center's collaborators is [Pia Puolakka](https://www.linkedin.com/in/pia-puolakka-8218a9b3/), the Finnish forensic psychologist and founder of [MindTech](https://www.mindtechfin.com/), who led Finland's Smart Prison Project: an operational model built around personal cell terminals, controlled digital access, education, rehabilitation, communication, digital and AI literacy, modernized offender management, and the ethical governance of artificial intelligence within custodial systems.
I did not go looking for any of this. It arrived as an adjacency. You find your name in a directory, you read down the list of people standing beside you, and one of them turns out to be the architect of a national program for placing a networked terminal in every cell of a prison. And then—because that is what a mind does when it has spent years watching how systems name themselves—you notice that the vocabulary of that program is not correctional vocabulary at all.
**Assessment. Need. Responsivity. Pathway. Dosage. Personalization. Controlled access. Earned progression. Absorption capacity. Graduated provisioning of privilege.**
That is the vocabulary of *underwriting*. It is spoken by mortgage servicers, catastrophe modelers, index administrators, rating analysts, sovereign-risk desks, and capital-flow management frameworks. It is the language of institutions whose function is neither to punish nor to heal, but to decide **how fast something should be permitted to move**.
Smart prisons are not the subject of this essay. They are the most legible *instance* of a subject otherwise nearly impossible to see: a distributed **risk-governance operating system** that ranks persons, properties, territories, institutions, and capital flows according to modeled future performance, and then differentially provisions their affordances. The prison is simply where the operations remain visible, because there the walls are still concrete and someone still has to write down a release date.
Everywhere else, the walls are rendered at runtime.
### II. The Second Adjacency: Eighty-Seven Million Proofs of Concept
Before the correctional machinery is described, the informational machinery has to be admitted, because it was demonstrated first, at scale, on the general population, and nobody had to be convicted of anything.
**Cambridge Analytica** is usually filed under *election scandal*. That filing is a category error. What that operation actually demonstrated — publicly, forensically, and in front of two national legislatures — was the complete loop this essay is about, running on the **informational substrate** instead of the correctional one.
Reconstruct it as a control system rather than as a news story. First, a **public research layer** builds the eye: academic psychometrics at Cambridge established that ordinary Facebook activity — Likes, nothing more — could be used to infer personality traits, dispositions, and vulnerabilities with usable accuracy. Then a **harvesting layer** resolves identity: a personality app collected data from a few hundred thousand consenting users and, through the platform's friends-permissions architecture, reached roughly **87 million profiles** of people who had consented to nothing. Then an **inference layer** converts observation into a proprietary verdict: psychographic segments, matched against voter files. Then a **provisioning layer** acts: different people receive different informational worlds, targeted to their inferred susceptibilities. And then the **response is measured** and fed back.
**Observation → identity resolution → feature construction → score → differential provisioning of affordance → measured response → updated score.**
That is not an analogy for the risk-and-needs architecture. It is the *same architecture*, and it was deployed against people who were not offenders, not supervised, not classified, not adjudicated — merely present.
The consequences were financial and reputational rather than penal: the Federal Trade Commission imposed a **\$5 billion** penalty on Facebook in 2019, the SEC added \$100 million, the UK regulator fined the company the statutory maximum available to it, and Cambridge Analytica dissolved. Note what was *not* dismantled: the psychometric science, the identity-resolution stack, the inference layer, the targeting infrastructure, or the business model. **The covenant — consent, disclosure, the platform's own API permissions — was revoked. The sensorium was retained.** This will turn out to be the general law of every system in this essay.
And here the adjacency turns personal, because the man whose company built that inference layer is also, with his wife, the largest private funder in America of the movement to **reduce incarceration** — and the funder of a research program to embed **thousands of sensors in human tissue**.
That is not a leap. It is a **topology**, and it is the reason this essay exists. Hold the three vertices in view: an **informational-affordance layer** capable of expanding and contracting what a person sees and can act on; a **biological-measurement layer** being built to make the interior of the body continuously legible; and a **justice doctrine** whose explicit purpose is to move correction *out of the building* and into open society. None of the three is a prison. Assembled, they are the complete component set for a perimeter that does not require one.
### III. Finland: The Cell That Learns
Begin the correctional half at the humane pole, honestly stated.
The **Smart Prison** project ran from 2018 to 2022 inside the Finnish Criminal Sanctions Agency, managed by Pia Puolakka, a forensic psychologist in the Finnish Prison and Probation Service since 2012. Its concrete deliverable was the new women's prison at **Hämeenlinna**, opened in October 2020, with a self-service terminal in every cell — education, rehabilitative content, family contact, public services, and a whitelisted portion of the network. In 2019, Finnish prisoners began working through the University of Helsinki course *Elements of AI*, delivered through the same restricted channel.
The proposition is unimpeachable and correct: **a person who must return to a digitally saturated society cannot be rehabilitated through enforced technological illiteracy.** Puolakka's formulation is not "technically enhanced custody." It is **digital prison reform**.
But the decisive innovation was never the laptop. It was the conversion of custody into an **adaptive interface**. The institution determines what information reaches the person, which choices are presented, which behaviors are recorded, which competencies are cultivated, and which pathways toward reintegration become visible at all. What correctional language calls **controlled digital access** is, computationally, a bounded personalization environment.
The cell stops being a room containing a computer. It becomes a **responsive cognitive perimeter** — an environment that learns enough about its occupant to decide which version of the world the occupant is permitted to encounter. Which is precisely what a personalized feed is, and precisely what Cambridge Analytica proved could be built for anyone.
The confined person must become literate enough to operate the system; the system becomes vastly more literate about the confined person. **AI literacy**, under custodial conditions, has two faces: learning to use computational systems, and learning how little of their operation is visible from inside the experience they construct.
Finland knows this. Puolakka served from 2021 to 2023 on the Council of Europe expert group that drafted **[Recommendation CM/Rec(2024)5](https://rm.coe.int/cm-rec-2024-5/1680b1fec1)**, adopted 9 October 2024, governing the ethical use of artificial intelligence by prison and probation services. Hold that instrument. It is the road not taken.
### IV. Huxley's Objection: Causal Plurality and Epistemic Enclosure
John Gannon's interest in Aldous Huxley sits inside this network like a splinter, and it belongs at the front of the argument.
Huxley's observation is that superficially identical behavior can arise from radically different causal histories. A missed appointment may indicate defiance, illness, an employment conflict, a caregiving emergency, a broken-down car, homelessness, confusion, or a clerical error made by the institution itself. The behavioral trace is one datum. The causal manifold behind it is irreducible.
Every actuarial system must nonetheless **compress causal plurality into administratively actionable categories**. That compression is the source of the system's power and of its characteristic injustice.
Where it goes uncorrected, personalization becomes **epistemic enclosure**:
**the model curates the environment according to prior inference → the subject responds to the curated environment → the response becomes new evidence for the prior inference → the system progressively validates itself.**
This is the closed recursive loop from which the invisible prison derives its most refined power. No malice is required to build it. Sincerity is sufficient. Every participant may be helping.
Huxley supplied the political intuition as well: **social control need not appear as terror.** It can be embedded in pleasure, convenience, normalization, therapeutic vocabulary, managed information, and the gradual harmonization of the person with an administered environment. What machine learning adds is the capacity to stop broadcasting a common ideology and to generate instead a **distinct corridor for each individual**. Mass control becomes individually rendered. The prison has no single wall because every subject receives a bespoke perimeter.
## PART TWO — GENEALOGY OF THE ADAPTIVE PERIMETER
### V. 1964: The Tether Was Invented as a Reward
The ankle monitor did not begin as a punishment. This matters more than almost anything else in this history, and it is nearly always forgotten.
At Harvard in the early 1960s, Ralph and Robert Schwitzgebel ran the **Streetcorner Research** project, building portable electronic transceivers — assembled partly from surplus missile-tracking hardware — to maintain contact with young offenders in Cambridge, Massachusetts. Their frame was explicitly Skinnerian. The device was conceived within the logic of **operant conditioning**: it was to transmit signals *to* the wearer, delivering positive reinforcement for prosocial behavior. Robert Gable described the original aim decades later as rewarding noncriminal conduct rather than punishing offenders — **electronic rehabilitation**, not electronic imprisonment.
The rehabilitative payload was stripped out on the way to market. In the late 1970s a New Mexico district judge, **Jack Love**, took the idea from a *Spider-Man* strip in which the Kingpin fixes a tracking bracelet to the hero, and worked with an engineer to build a working device; the first monitored subjects appeared in Albuquerque in 1983. Love, having tested the unit on himself, described the sensation in five words that outlive every euphemism since: it put him on a very short leash. The commercial product was sold as an **electronic handcuff**.
So the genealogy of the invisible prison runs: **a reward channel is invented; the reward channel is inverted into a restraint channel; and the restraint channel is re-narrated, forty years later, as a rehabilitative affordance.** Every instrument in this essay repeats that cycle. The tablet in the Finnish cell is a reward channel. The tablet in the American cell is a billing channel. The difference is not the hardware.
Alongside the hardware came the intellectual frame. In 1992, **Malcolm Feeley and Jonathan Simon** named the **New Penology**: a shift away from the individual offender, his guilt and his transformation, toward the **actuarial management of aggregates**. The New Penology does not ask what a person deserves. It asks what a category costs. Once that question is legitimate, everything downstream is not merely possible but rational.
### VI. 2004: The Results-Driven Organization
Beginning around **2000**, the Administrative Office of the United States Courts engaged consultants — led by **PricewaterhouseCoopers** and **IBM Business Consulting Services**, during the period in which PwC's consulting arm passed into IBM — to conduct a comprehensive **Strategic Assessment of the Federal Probation and Pretrial Services System**. The resulting 2004 report did not concern penology. It concerned **management**. Its central recommendation was that federal probation become a **results-driven organization** with a comprehensive outcome-measurement system.
That is the moment the machine's grammar was installed. Not by a criminologist. By a **business-process engineer**.
What emerged operationally was the **Post-Conviction Risk Assessment (PCRA)** — superseding the older Risk Prediction Index — and the doctrine of **Risk-Need-Responsivity**:
**observe → classify → diagnose → prescribe → monitor → reassess.**
PCRA does not merely calculate static danger. It fuses officer observation with subject self-report across employment, substance use, social networks, cognition, and attitudes, then recalculates those **dynamic criminogenic needs** at intervals. That is not a sentence. It is a **control loop with a human being inside it**.
Note what a *dynamic* factor is in engineering terms. A static factor — age at first arrest, prior convictions — is a constant. A dynamic factor is a **control surface**: a variable the institution can act upon, and whose movement it can then read back as evidence of its own efficacy. The invention of dynamic risk is the invention of the correctional actuator.
### VII. Dosage: When Time Stops Being the Unit of Punishment
The full implication only becomes visible in **dosage probation**, developed through the National Institute of Corrections with the Center for Effective Public Policy and the Carey Group.
The proposal is simple and radical: **the length of supervision should not be a period of time.** It should be the number of **hours of intervention** required to reduce assessed risk. A moderate-to-high-risk person might be assigned a target on the order of two hundred hours of cognitive-behavioral programming; the term ends when the dosage target is met and dynamic risk has fallen — not when the calendar says so.
Sit with what has happened. For two centuries the unit of punishment in the Anglo-American world was **time**: a determinate quantity, imposed by a court, in public, on the record, appealable, and **finite by arithmetic**. Dosage replaces it with **measured behavioral change**, assessed by an instrument, administered by a supervising officer, terminated when a model says risk has fallen far enough.
This is not necessarily crueler. Under dosage, the compliant person leaves early; under calendar sentencing, they cannot. That is the humane pole, and it is real.
But the deep structure is that **liberty has been converted from a right that returns automatically into a commodity earned against a threshold set by the state's own model** — a threshold the subject cannot audit, cannot appeal on its merits, and cannot see. The clock was stupid, but the clock was neutral. The model is intelligent, and it is not.
Everything in the financial half of this essay is a variation on that single substitution: **an obligation with a fixed term is replaced by an obligation with a modeled term.** That is what a callable bond is. That is what a dynamic insurance rate is. And it is now what a sentence is.
### VIII. Semantic Compliance: STARR, CNVC, and Lawful Self-Management
The loop needed an actuator, and federal probation built one.
**STARR** — Staff Training Aimed at Reducing Rearrest — trained officers to deploy cognitive-behavioral technique inside ordinary supervision contacts, paired with **motivational interviewing**, a method engineered to elicit *internally articulated* reasons for change while preserving the phenomenology of alliance. Officer interactions are coded into a tracking system. The Administrative Office later contracted the **University of Cincinnati Corrections Institute** to build the **Criminogenic Needs and Violence Curriculum**, integrated with PCRA. Federal probation publications by 2020 described these as one **integrated supervision framework**, and described motivational interviewing as a natural fit because it let officers combine **control with a working alliance**.
That phrase is the semantic key to the entire architecture, and the government wrote it about itself.
The Administrative Office's own guidance states the terminal objective plainly: the goal for each person under supervision is **lawful self-management** — making personal choices not to offend, eventually without the oversight of the justice system.
Read as an engineering specification rather than a moral aspiration, that says: **the system's objective is to install itself inside the subject.** Successful supervision is supervision that no longer requires a supervisor, because the supervisor has been internalized.
Which produces the phenomenon the architecture depends on and never names: **semantic compliance**. The subject learns which disclosures count as insight; which goals count as progress; which emotional expressions register as responsivity; which account of one's own past unlocks the next tier of affordance. The person is not merely required to behave. The person is required to **narrate themselves in the institution's language, and to be sincere while doing it.**
A subject producing the authorized account of their own transformation generates exactly the evidence the model was built to detect — which the model reads as confirmation that its intervention worked. **The instrument measures its own echo.** The same architecture now appears in employment coaching, platform moderation appeals, therapeutic case management, immigration credibility assessment, and every consumer AI system that rewards the user for phrasing a need the way the system can parse.
Huxley's objection, in operational terms: **the system cannot distinguish a transformed person from a fluent one.**
## PART THREE — THE ORGAN AND ITS SUNSETS
### IX. Whole-of-Government Reentry, 2011–2016
On **5 January 2011**, Attorney General Eric Holder convened the first meeting of the **Federal Interagency Reentry Council**. More than twenty federal agencies participated. Its premise was explicit: reentry is not a Department of Justice problem. The reentry population is already present in emergency rooms, homeless shelters, unemployment lines, child-support offices, veterans' hospitals, benefit systems, and schools.
Read that with a cybernetician's eye and something else appears. The Council converted a **judicial status into a cross-domain personalization problem**. A person's actionable world — employment, housing, health care, education, credit, identification, family reunification — could now be expanded or narrowed through a coordinated set of institutions, almost none of which look correctional from outside.
This was not a scheme. It was a humane and overdue repair. But it was also the first American structure capable of producing an **invisible rehabilitation state** without requiring any single participant to perceive the total design.
The hierarchy is **polycentric, not pyramidal**. Congress, DOJ, OJP, and the Bureau of Justice Assistance set grant conditions, evidence standards, and performance measures. The **Second Chance Act** supplies the statutory and funding spine. The **Smart Suite** — Smart Supervision, Smart Reentry, Smart Probation — supplies the branded instruments, alongside **Evidence-Based Decision Making** from the National Institute of Corrections. The **Justice Reinvestment Initiative** supplies the state-system redesign mechanism. The **National Reentry Resource Center**, run by the **Council of State Governments Justice Center**, supplies translation. **Policy Research Associates** supplies the **Sequential Intercept Model**. The **Center for Effective Public Policy**, the **Carey Group**, and **UCCI** supply doctrine, curricula, and training. States and counties implement. And at the human contact layer stand officers, case managers, clinicians, clergy, mentors, employers, landlords, and peer specialists.
The Bureau of Prisons' **Community Reentry Network** alone maintains a national database of roughly **5,900 vetted faith-based and community organizations**.
None of these people needs operational knowledge of an encompassing system. Each can understand themselves as mentoring, employing, housing, counseling, or gently redirecting one person. **The aggregate behaves coherently while its components remain morally sincere and informationally compartmentalized.**
This is where the proverb about good intentions stops being proverbial and becomes infrastructural.
In 2014, GAO named the Reentry Council one of four model interagency collaborations in the federal government. On **29 April 2016**, President Obama issued a Presidential Memorandum formally establishing it, and the Council — co-chaired by Attorney General Lynch and Domestic Policy Council Director Cecilia Muñoz — published *[A Record of Progress and a Roadmap for the Future](https://nationalreentryresourcecenter.org/resources/federal-interagency-reentry-council-record-progress-and-roadmap-future)*.
For five years, the coordination layer was **visible, named, chartered, audited, and findable**. It is the last time it will be true.
### X. Regime Transfer: The Office of American Innovation, the First Step Act, PATTERN
What happened next is usually narrated as destruction. It was **transfer**.
On **7 March 2018**, President Trump signed **[Executive Order 13826](https://www.federalregister.gov/documents/2018/03/12/2018-05113/federal-interagency-council-on-crime-prevention-and-improving-reentry)**, establishing the **Federal Interagency Council on Crime Prevention and Improving Reentry**, co-chaired by the Attorney General, the Assistant to the President for Domestic Policy, and the Senior Advisor heading the **White House Office of American Innovation** — Jared Kushner's office. **Section 5 revoked Obama's April 2016 Presidential Memorandum.**
The organ persisted. The charter, the custodians, and the moral wrapper were replaced. That the successor body sat partly inside the Office of American Innovation is not decorative: that office approached government through business-process redesign, private-sector partnership, metrics, and outcome optimization — **precisely the managerial ontology PwC and IBM had installed in federal probation eighteen years earlier**. The political aesthetics inverted. The engineering did not.
Nine months later the **First Step Act** (P.L. 115-391, 21 December 2018) completed the design. It reauthorized the Second Chance Act. It required the Attorney General to build a **risk and needs assessment system** covering the entire federal prison population. It tied programming, **earned time credits**, and prerelease custody to modeled recidivism risk. Its own text specifies that **increasingly less restrictive conditions** should be imposed on those who demonstrate continued compliance — an admirably candid description of a graduated, feedback-driven perimeter. And in **§505, "Federal interagency reentry coordination,"** Congress made interagency reentry coordination a **statutory duty of the Attorney General**, with a report due within two years.
The instrument was **PATTERN**. By January 2020, every federal prisoner had received an initial assessment. What PCRA had been for community supervision, PATTERN became for custody itself. The President called it **"smart confinement."**
Now the asymmetry, which is the deepest political finding here. The same administration curtailed police consent decrees, rescinded *Smart on Crime*, reversed the Yates memorandum on private prisons, and constrained pattern-or-practice civil-rights enforcement. Every one of those mechanisms was a **supervisory loop pointed upward, at the state**: assess the department, classify the deficiency, prescribe correction, monitor, reassess. Identical cybernetic grammar to PCRA and PATTERN, aimed at institutions rather than persons.
**The upward feedback channel was cut. The downward control channel was extended.**
Not the destruction of the adaptive supervision state. The removal of its counterweight.
### XI. The Fuse — and the Convergence of January 2021
Here is the fact that reorganizes the entire period, sitting in plain text in the Federal Register.
**Section 6 of Executive Order 13826 was a termination clause.** The order, and the Council it established, "shall terminate 3 years after the date of this order." Not upon revocation. Not at a successor's pleasure. By its own text, on **7 March 2021**, the Federal Interagency Council on Crime Prevention and Improving Reentry **ceased to exist**.
Nobody killed it. It was built with a fuse.
And now place beside that date a second one. On **27 January 2021** — six weeks earlier — the **Chan Zuckerberg Initiative**, by then among the largest private funders of criminal-justice reform in the United States, announced that it was **divesting its criminal-justice grantmaking**: \$350 million over five years to a spun-off entity (the Justice Accelerator Fund, later **The Just Trust**), \$100 million to **FWD.us**, and a stated intention to refocus CZI on science, education, and community. The advocacy portfolio was handed out. The commitments were later fulfilled and the social-advocacy funding wound down.
So within a single quarter, the **public coordinating covenant** for American reentry expired by operation of its own clause, and the **largest private covenant** in the same field was packed up and handed to somebody else.
Nothing coordinated the two. That is the point. Two entirely independent institutions — one a federal executive body, one a family philanthropy — shed the **covenant layer** in the same six weeks, for their own reasons, while the **measurement layer beneath both** continued to compound without interruption. PATTERN kept scoring. The consumer files kept accumulating. And the same philanthropy that had just divested justice reform would, twenty-five months later, announce a program to embed **thousands of sensors in human tissue** in Chicago.
**The covenant is temporary. The ledger is durable. The institution stops helping before it stops measuring.**
Meanwhile the statutory duty under First Step Act §505 sat unexecuted. The Attorney General finally convened the **Reentry Coordination Council** in **October 2021** — nearly a year past the deadline — with HUD, Labor, Education, HHS, VA, and Agriculture. Its report, *[Coordination to Reduce Barriers to Reentry](https://www.justice.gov/archives/opa/press-release/file/1497911/dl)*, landed in **April 2022**. One month later, **Executive Order 14074** (25 May 2022) established yet another body — the **Alternatives and Reentry Committee** — producing the **Alternatives, Rehabilitation, and Reentry Strategic Plan** on 28 April 2023. In January 2025, EO 14074 was revoked.
Count the names. **Federal Interagency Reentry Council. Federal Interagency Council on Crime Prevention and Improving Reentry. Reentry Coordination Council. Alternatives and Reentry Committee.** Four names for one function in twelve years, each anchored to a different legal instrument, each generating a distinct and non-overlapping documentary record, each becoming a **dead search string** the moment its charter lapsed.
This is the mechanism of invisibility, and it is not concealment. It is **differential decay**.
### XII. The Pandemic: The House Becomes a Facility
The fuse burned down at the precise moment the perimeter had become more plastic than at any point in modern American custodial history.
In late March 2020, under **§12003(b)(2) of the CARES Act**, the Attorney General directed the Bureau of Prisons to move as many people as possible out of secure facilities into **home confinement**. In the same statute, **§12003(c)** mandated free video and telephone contact for federal prisoners for the duration of the emergency.
In a single week of legislating, the United States both **dissolved the wall** and **statutorily digitized contact across it**. This is the American smart-prison moment. It was never called that, because it arrived as public-health emergency management rather than correctional philosophy.
Then the boundary oscillated. On **15 January 2021**, the Office of Legal Counsel opined that BOP would be *required* to recall every CARES Act home-confinement prisoner when the emergency ended. On **21 December 2021**, OLC reversed itself. The **final rule** published 4 April 2023 and took effect 4 May 2023. As of 23 January 2023, **3,434 people** were in CARES Act home confinement, of whom 2,026 had more than a year remaining. The rulemaking stated the economics without embarrassment: roughly **\$120 per day** in a facility, roughly **\$55 per day** at home.
For twenty-three months, several thousand Americans lived in their own houses under a legal status the federal government could not decide the meaning of. They worked. They raised children. They were monitored, tested, visited, and reported on.
And when one of them — pulled back inside from home confinement — argued she had been deprived of liberty without due process, the government's answer was the most important sentence in this investigation: it contended that home confinement is **merely another facility designation** under 18 U.S.C. § 3621, no different in kind from a transfer between institutions. Practitioners glossed the position, drily, as *FCI Home*.
There it is, in a brief, in open court.
**The house is a facility. The perimeter is an administrative attribute. The prison is a designation, not a place.**
You do not need to allege an invisible prison. The United States has already asserted one, as a defense.
## PART FOUR — THE CIRCULATORY SYSTEM
*Everything above names the organs. What follows is how blood moves between them.*
### XIII. Identity Resolution: How the Scored Object Is Assembled
Before anything can be scored, **the thing to be scored must be manufactured**.
A person does not arrive at an institution as a unified object. They arrive as **fragments**: a booking record with one spelling, a credit file with another, a lease under a married name, a utility account at a prior address, a phone number that used to belong to someone else, a device identifier, a payroll entry, a patient index, a benefits case number. Nothing joins automatically. The joining is a manufactured achievement, and it has a name: **entity resolution**.
The techniques are unglamorous and decisive: **master data management, master person indexes, deterministic and probabilistic record linkage, fuzzy matching, address normalization, householding, device and account graphing, KYC and AML customer-identification programs, identity proofing**. Each makes a bet — that these two records refer to the same human being — and hands downstream systems a **persistent subject** to act upon.
The correctional systems say so explicitly. Illinois built **Offender-360** on a commercial customer-relationship-management platform, and the stated ambition was a **single view of the offender**. Read that phrase again. It is CRM language — what a bank says about a customer. The person under supervision has been rendered as an *account*.
And this is precisely what **MERS** did for the mortgage: a registry against which servicing rights and beneficial interests can be tracked and transferred without repeated local paper assignments. It is entity resolution for financial claims. The **master person index** is entity resolution for people. And the friends-permissions graph that produced 87 million Cambridge Analytica profiles was entity resolution for the social world.
**No score can travel until an identity is resolved. Identity resolution is therefore the true perimeter fence, and it is built out of string-matching heuristics, address-normalization rules, and probabilistic thresholds no subject has ever seen.**
Two consequences follow, both severe. A false match — the wrong criminal record joined to the right name — propagates instantly and everywhere, because the entire value of the resolution layer is that downstream users **do not re-derive it**. And the subject cannot contest the resolution, because the subject does not know it happened. **You cannot appeal a join.**
### XIV. The Consumer-Reporting Spine: Where the Person Becomes Portable
If ICE and MERS are the portability spine for **property claims**, the **consumer-reporting and data-broker industry** is the portability spine for **person claims**. This is where the invisible prison acquires nationwide reach without a single federal database.
**Credit bureaus** — Equifax, Experian, TransUnion — hold payment histories and, through workforce, tenant, and insurance-scoring divisions, extend into employment verification, rental admissibility, and premium setting. **LexisNexis Risk Solutions** and comparable public-record aggregators join criminal records, court filings, addresses, relatives, associates, vehicles, and identity signals into one retrievable profile. **CoreLogic** holds property, title, and tenant data. **FICO** and its peers supply the scoring layer. **Background-screening firms** convert court records into employment decisions. **Fraud and identity networks** correlate devices and accounts across institutions.
Each is individually lawful. Together they are the machine by which **a person can be down-ranked across every domain of life without any government computing a single social score**. The eviction filing that was dismissed still appears in the tenant screen. The arrest that produced no conviction still appears in the background check. No leap is required — only an aggregator, a customer, and an API.
Note what this does to the correctional architecture. PATTERN and PCRA are *internal* instruments; their formal jurisdiction ends at the agency door. But the **event stream they generate** — supervision status, noncompliance, violations, revocations, and the court records beneath them — flows into a commercial layer with **no jurisdictional limit at all**. The state releases the person. The record does not.
**The person walks out of the building. The claim about the person keeps working.**
### XV. Report Routing: How a Neighbor's Perception Becomes an Administrative Fact
The American reentry ecology is capable of generating something no agency could afford: an **ambient human sensorium** around a classified population.
The doctrine is not hidden. It is in the volunteer handbooks. A juvenile-justice technical guide describes the probation officer as the court's **eyes and ears**. Idaho's mentoring program tells volunteers, in its recruitment material, that they will be **another set of eyes and ears**, working in concert with the reentry plan established by the case manager and supervising officer. And the international proof of concept is unambiguous: Japan's ***hogoshi*** system fields roughly **47,000 volunteer probation officers**, embedded in the neighborhoods of the people they supervise.
But perception does not become power until it has a channel. Here is the routing layer — the connective tissue of the whole system.
**Formal reports** are structured, dated, permanent: probation case notes, collateral-contact verifications with employers and landlords, treatment-provider updates, urinalysis results, monitoring alerts, incident reports, suspicious-activity reports routed to fusion centers, child-welfare referrals, school reports, wellness checks. These enter case-management systems directly and are legible to the risk instrument.
**Semi-formal referrals** connect agencies with no shared jurisdiction: coordinated-entry and homeless-management information systems, multidisciplinary-team records, **behavioral threat assessment** team files, reentry task-force minutes, memoranda of understanding and data-sharing agreements that let a housing agency and a probation office see the same person. This layer is where **cross-domain fusion actually occurs**, and it is almost entirely invisible to the public.
**Informal reputational propagation** has no form at all: a mentor's remark, a landlord's hesitation, an employer's decision not to call back. Unrecorded, unappealable, frequently decisive.
These three registers are **asymmetrically coupled**. Information flows *upward* — from informal perception into semi-formal referral into formal record — with almost no friction, because every professional in the chain is trained to document concern. It does not flow *downward*. The subject cannot see the case note. The volunteer cannot see the risk tier. The landlord never learns what their hesitation triggered.
**The sensorium is one-way glass, and everyone standing at it believes they are simply looking out a window.**
### XVI. Negative-Data Asymmetry: The System's Differential Memory
Here is the property that makes the perimeter progressively harder to escape, and it is not a scoring property. It is a **memory** property.
**Adverse information is structured, standardized, transferable, and durable. Exculpatory information is narrative, local, unstructured, and perishable.**
An arrest generates a record even without conviction, and that record enters commercial aggregation. A dismissed eviction filing still surfaces in tenant screening. A missed payment survives the hardship that caused it. A supervision violation is coded, timestamped, permanent; the reason for it — the shift that ran late, the child who was sick, the bus that never came — lives, if anywhere, in a case note no downstream system will ever read.
The result is a system with **differential memory**. It remembers what you did wrong in a portable format and what you explained in a format that does not travel. Over time this thickens the adverse profile relative to the mitigating one, at every institution, simultaneously, forever.
This is why "just correct the error" is not an answer. The error is usually not a false statement of fact. The error is that **the true fact has been separated from its cause and given a passport**, while the cause was left at home without one.
Huxley's causal plurality is not merely *compressed* by the actuarial system. It is **format-discriminated**. The system did not decide that the reason doesn't matter. It decided that the reason doesn't **serialize**.
### XVII. Time: The Expiration of Claims About Persons
Every score has a hidden temporal architecture, almost never discussed: **lookback windows, recency weighting, decay functions, retention schedules, model retraining intervals, expungement and sealing rules, certificates of rehabilitation, registry durations**, and the discoverability of the underlying records after the score itself has improved.
That last clause contains the trap. A risk score can improve while the records that produced it remain **indefinitely discoverable**. The state may forgive; the aggregator does not. A person can complete supervision, satisfy the model, earn the credits, receive the certificate — and still be denied the apartment, because the tenant screen does not run PATTERN. It runs the court record, and the court record has no expiry.
So the ethical demand is larger than the one usually made of algorithmic systems. It is not enough to demand **decommissioning of the model**. What is required is the **expiration of the claim about the person**. Without temporal limits on the propagation of adverse records, rehabilitation is promised rhetorically while **informational punishment becomes permanent**.
Dosage removed the calendar from the front end of punishment. Perpetual record discoverability removed it from the back end. What remains is a punishment with **no arithmetic termination in either direction**.
### XVIII. Ranking, Eligibility, Pricing, Intervention: Where Inference Becomes Coercion
"Social credit" becomes an unfalsifiable accusation unless the mechanism is specified. So specify it. There are **four distinct layers**.
**Ranking** orders subjects. By itself it changes nothing; a list is not a sanction.
**Eligibility** converts rank into a binary: admitted or excluded.
**Pricing** converts rank into a continuous cost: premium, rate, deposit, fee, verification burden.
**Intervention** converts rank into an action taken *upon* the subject: supervision dosage, mandated treatment, curfew, monitoring, home visit, restricted association.
Each is defensible in isolation. The invisible prison emerges **only where they become coupled** — and coupled is what they now are. The complete chain:
**observation → identity resolution → feature construction → model → score → rank → eligibility or price → intervention → observed response → updated score.**
That is the circulatory system, and it identifies the precise coordinate at which inference becomes coercion: **the coupling point between rank and provisioning.** Everything before it is epistemology. Everything after it is power. The transition is invisible because no single institution owns both sides of it — the modeler does not provision, the provisioner does not model, and each can therefore describe itself, accurately, as merely doing its job.
## PART FIVE — CHICAGO
### XIX. The Instrumented City
If you want to see every layer of this architecture being built simultaneously, in one metropolitan area, look at Chicago. Not because Chicago is uniquely culpable — because Chicago is uniquely **legible**. Four scales of the same sensorium are under construction there at once, funded by four different kinds of money, and nobody has ever drawn them on one map.
**The behavioral scale.** From 2012, the Chicago Police Department maintained the **Strategic Subject List** — the "heat list" — a person-based predictive-policing instrument that ultimately assigned risk scores to roughly **398,684 individuals**, on a scale running past 500. Its stated uses were to connect people to social services and to serve as an investigative resource. Its operational use was the **custom notification**: police, social workers, and community leaders arriving at the door of a person who had been convicted of nothing, to warn and to help, simultaneously. Chicago's Office of Inspector General raised concerns about its efficacy, and the list was quietly decommissioned in **November 2019**. RAND's evaluation found that the listed group was no more likely to become a homicide *victim* than a comparison group — but was more likely to be **arrested** for a shooting. The list did not predict violence. It predicted police attention, and then produced it.
**The urban scale.** The **Array of Things** — the University of Chicago and Argonne National Laboratory's networked urban sensing project — installed instrumented nodes on city infrastructure to measure the environment continuously: a fitness tracker for the city. Alongside it, acoustic gunshot-detection sensing operated across the South and West Sides until the city let the contract lapse in 2024.
**The administrative scale.** The Illinois Department of Corrections built **Offender-360** on a commercial CRM platform to achieve a **single view of the offender**. Cook County adopted algorithmic pretrial risk assessment. And on **18 September 2023**, Illinois became the first state in the country to **abolish cash bail entirely** under the Pretrial Fairness Act — replacing a monetary price on pretrial liberty with a **judicial determination of dangerousness and flight risk**, informed by assessment. Consider what that substitution is, structurally, and hold your applause and your objection at once: the *price* of freedom, which was cruel and regressive and unjust, was replaced by a *model* of freedom, which is none of those things and is also unappealable in a way a bail schedule never was. The covenant layer was removed. The inference layer took its place. Illinois did the right thing and, in doing it, completed the architecture.
**The cellular scale.** On **2 March 2023**, the **Chan Zuckerberg Initiative** announced the **CZ Biohub Chicago** — the first expansion of the Biohub Network outside California — uniting the **University of Chicago, Northwestern University, and the University of Illinois Urbana-Champaign**, led by Shana O. Kelley, and formally launched that October in Fulton Market with the Governor of Illinois and Priscilla Chan present. Its declared method: **embedding thousands of sensors and sampling probes in tissues** to monitor molecular and cellular signals at unprecedented resolution, in service of measuring and ultimately treating **inflammation** — and, in CZI's standing formulation, of the goal to **cure, prevent, or manage all disease by the end of the century.** Illinois publicity called the object of the work an **instrumented tissue mimic**.
Four scales. **The block. The body. The record. The cell.** All being made continuously measurable, in one city, in the same decade, by institutions that have never sat at the same table.
### XX. Robert McDaniel's Front Door
Everything in this essay happened, once, in a doorway in the Austin neighborhood on the West Side of Chicago, in the middle of 2013, to a twenty-two-year-old man named **Robert McDaniel**.
He had no violent criminal history — possession and gambling. He answered a knock and found a **cohort** on his porch: two uniformed police officers, **a neighbor working with the police**, and a social worker. The social worker offered help — a job, mental-health services. The police offered a warning: from now on, the department would be watching him. An algorithm had determined that he was more likely than **99.9 percent** of Chicago's population to be shot, or to have a shooting connected to him.
Read the composition of that porch again, because it is the entire thesis rendered as a tableau. **The score. The intervention team. The clinical offer. The custodial warning. The ambient human sensor — a neighbor, morally sincere, informationally compartmentalized, standing on the step.** *Control with a working alliance*, delivered to a private residence, without charge, without conviction, without a forum in which to contest the classification, because formally nothing had been done to him. He had been *helped*.
Then the loop closed. Police lingered around the store where he worked. They visited. They searched. They came back. And his neighbors, watching officers arrive at his house repeatedly without ever arresting him, drew the only inference available in that environment: **he must be an informant.** He was shot. Later, he was shot again.
The model predicted that Robert McDaniel would be involved in a shooting. The model's own intervention **produced the shooting**. And the outcome then stands, in the data, as vindication of the score.
This is not a metaphor for epistemic enclosure. It is the thing itself, with a case number. **Adverse classification → intervention → destroyed social standing → adverse outcome → confirmation of the classification.** The system did not fail. It functioned exactly as designed, and the design is the failure.
Everything else in this essay — dosage, PATTERN, MERS, the tenant screen, the catastrophe model, the risk tier, the earned credit — is that porch, at a different scale, with the humans removed.
### XXI. The Measurement Layer: CZI, the Healable Person, and the Doctrine of Non-Incarceration
Now the connection that the record actually supports, stated with precision, because it is more consequential than the accusation people expect.
**There is no procurement record, grant, or contract linking the Chan Zuckerberg Initiative, the Biohub Network, or Meta to the design or operation of community custody, electronic monitoring, risk scoring, or correctional technology. I assert none.** Every documented movement runs the *other* way: CZI's justice portfolio, before it was divested, funded **decarceration** — clean-slate record clearing, prosecutorial accountability, bail reform, reentry networks — and its founders' interest reportedly began with a 2015 visit to San Quentin to see a coding class taught to incarcerated men.
That is exactly why it belongs here.
**Reducing incarceration means dealing with criminality outside the building.** This is not a hidden agenda; it is the stated, bipartisan, foundation-supported doctrine of the last fifteen years, and it is correct. But it has an infrastructural entailment that almost nobody says out loud: if the person is not in a cell, then **the functions of the cell — observation, classification, restriction, intervention, compliance measurement, graduated release of privilege — must be discharged somewhere else.** They must be discharged in housing, employment, treatment, credit, mobility, relationship, and information. Which means decarceration does not eliminate the perimeter. **It distributes it.**
And a distributed perimeter requires exactly three things: an **information layer** that can modulate what a person encounters; a **measurement layer** that can render the interior of a person legible; and a **doctrine** that reframes deviance as a condition to be healed rather than an act to be punished.
Trace the three, without inventing a single connection:
**The information layer** is the one the Cambridge Analytica affair proved operational — psychographic inference from ordinary behavioral exhaust, coupled to differential provisioning of what a person sees. On the platform side, this exists today as ranking, flagging, redirection, and content intervention. Damping and acceleration, applied to the epistemic environment.
**The measurement layer** is CZI Science. Its head of science from 2016 to 2022, **Cori Bargmann**, co-chaired the NIH working group that set the strategy for the **BRAIN Initiative** — map cell types, then circuits, then build tools to intervene. CZI co-funded the **Allen Institute's** work on molecularly defined cortical cell types within the **Human Cell Atlas**, and supported the spatial-transcriptomics consortia that made single-cell mapping tractable. The Allen Institute in turn played a central role in **MICrONS**, the IARPA-funded reconstruction of a cubic millimeter of mouse visual cortex — some **200,000 cells and 523 million connections** — whose declared purpose is to **reverse-engineer the algorithms of the brain** for machine learning. And CZ Biohub Chicago is now building the tissue-scale instrument: thousands of embedded sensors, measuring inflammation in living tissue, on a fifteen-year horizon.
**The doctrine** is the one the reentry apparatus has been converging on for two decades: that violence, addiction, and supervision failure are **outcomes of trauma, stress, environment, and social determinants of health**. That doctrine is humane, evidence-supported, and — this is the part that matters — **it makes correction a medical rather than a juridical operation.** Once deviance is a health outcome, the entire apparatus of measurement, personalization, dosage, adherence, and whole-person case management applies to it *by right*. You do not need a warrant to help someone. You do not need a conviction to treat someone. You do not need a trial to intervene on a risk factor.
Now hold the three together. If disease is defined — as Biohub Chicago defines it — as **a disruption in molecular signals measurable by embedded sensors**, and if psychological and behavioral conditions are understood as arising from inflammation, chronic threat response, intergenerational trauma, and cognitive rigidity, then **conditions that present as behavior become legible to the same instrument.** Not by leap. By definition.
I want to be exact about the epistemic status of that sentence, because the whole argument depends on it. It is **not** a claim that CZI intends to apply biomedical tools to criminality, or that Biohub Chicago is a correctional program, or that anyone involved has entertained the thought. The public record contradicts each of those claims, and I am not making them. It is a claim about **structure**: that the doctrine of non-incarceration generates a demand for a distributed perimeter; that a distributed perimeter requires an ambient measurement layer; and that the measurement layer is being built — openly, philanthropically, with excellent intentions and Nobel-adjacent science — by the same institutional family that funded the doctrine.
**The cell that learns, in Finland, is a room. The cell that learns, in Chicago, is a cell.**
That is the sentence this entire essay was written to earn, and it is not an accusation. It is a description of what happens when the humane project of emptying prisons meets the technical project of making everything measurable, and nobody is assigned to notice that they are the same project viewed from two ends.
## PART SIX — THE FINANCIAL HOMOLOGUE
### XXII. Governed Flow: The IMF and Absorption Capacity
In May 2012 the IMF published Working Paper 12/138: **Leonor Keller and Ibrahim Chowdhury, *[Managing Large-Scale Capital Inflows: The Case of the Czech Republic, Poland and Romania](https://www.imf.org/external/pubs/ft/wp/2012/wp12138.pdf)***. Its subject is the governance of **flow**: what a state should do when mobile capital arrives faster than its institutions can safely metabolize it. It evaluates macroeconomic instruments against **capital flow management measures** — the technical euphemism for a throttle.
Its governing concept is stated plainly: coping with large inflows is **less challenging for countries with strong absorption capacities**, and a country's capacity to absorb foreign capital depends on the depth and efficiency of its financial system.
**Absorption capacity.** How much flow can this entity take before the flow itself becomes damage?
That is not economics. It is **control engineering applied to a governed subject**:
A **country** is assessed for capacity to absorb capital. A **borrower**, for capacity to absorb credit. A **property**, for capacity to preserve value under hazard. A **municipality**, for capacity to absorb climate exposure without default. A **prisoner or supervisee**, for capacity to absorb **liberty**.
Dosage probation is a capital flow management measure applied to freedom. Two hundred hours of intervention is an **absorption-capacity threshold**. The IMF paper is not about prisons; it is about the same thing prisons are about — **who is permitted to move, how fast, and on whose model of their capacity to bear it.**
### XXIII. Poland: Where the Model Is Built, Far From Where It Lands
**ING Hubs Poland** operates as one of the group's global capability centres in **Katowice and Warsaw**, supplying technological and operational services across the bank, with declared concentrations in cybersecurity, financial and non-financial risk, compliance, and **model development, monitoring, validation, and governance**.
This is not evidence that a Dutch bank participates in American corrections. The importance is **morphological**. The same transnational group that originates, owns, services, or finances property claims maintains centralized hubs where the models governing those claims are **built, validated, and governed for worldwide deployment**. Katowice and Warsaw function as part of a distributed **epistemic factory**: teams constructing the systems that determine risk across multiple national markets, without being anywhere near the properties, borrowers, municipalities, or persons whose lives those systems modulate.
The circulation of personnel completes the picture. A career such as that of **Katarzyna Zajdel-Kurowska** — the Polish Ministry of Finance, the National Bank of Poland, representation at the IMF, service as a World Bank executive director, a Polish bank supervisory board, and leadership within the World Bank Treasury's partnerships and advisory operations — illustrates the ordinary, lawful movement among **finance ministries, central-bank reserve management, multilateral institutions, commercial-bank supervision, and global asset management**.
Below them sit the integrators: **Infosys, Accenture, Deloitte, EY, KPMG, PwC Polska**. No documentary evidence establishes that any of them designed an American correctional program, and I do not assert it. The defensible connection is functional and sufficient: **risk classification becomes governance only when it becomes interoperable, auditable, and reproducible across jurisdictions — and that is what integrators are for.**
The property and the person remain local. **The model governing them becomes transnational.**
### XXIV. ICE, NYSE, MERS: Making the Claim Portable While the Life Stays Still
**Intercontinental Exchange** is no longer merely the parent of the **New York Stock Exchange**. It spans exchange and derivatives infrastructure, **benchmark administration** through ICE Data Indices, fixed-income and municipal analytics, **climate-risk and property intelligence**, mortgage origination technology, servicing infrastructure, electronic recording, and mortgage-rights registration.
The pivotal acquisition is **MERS**. ICE took a majority equity position in **MERSCORP Holdings** in 2016 and, on **3 October 2018**, completed the purchase of all remaining interests. MERSCORP operates the **MERS System**, the national electronic registry tracking changes in **servicing rights and beneficial ownership interests** in U.S. residential mortgage loans across more than five thousand member institutions — allowing those interests to move within a national registry while MERS remains mortgagee or nominee in local land records, eliminating what the industry calls breaks in the chain of title.
ICE then assembled the rest of the stack: **Simplifile** for electronic recording, **Ellie Mae** for origination, and **Black Knight**, whose acquisition the Federal Trade Commission challenged; the settlement required divestiture of Empower and Optimal Blue, and ICE completed the remainder in 2023. ICE now markets property-level intelligence covering effectively the whole U.S. housing stock, and climate-risk analytics joined to municipal-debt exposure across most of the U.S. municipal market.
What that assembly accomplishes, as a single asymmetry:
**Land and its occupants are immobile. Financial claims upon land became electronically mobile.**
The household experiences one home and one obligation. The capital system experiences a **portable, data-bearing instrument** whose servicing, beneficial interest, valuation, insurance exposure, climate exposure, payment history, and portfolio role can be transferred, pooled, analyzed, securitized, and repriced — continuously, without the household's participation or knowledge.
Now read it with a person in place of a house. **The subject remains singular. The claim about the subject becomes portable.**
MERS did it for the mortgage; the master person index and the consumer-reporting spine did it for the human being; the social graph did it for the electorate. The innovation is never surveillance. The innovation is **transferability of the claim in the absence of the claimed**.
And the corollary, taught by American foreclosure history: the faster a claim travels, the harder it becomes to **contest its provenance in any downstream forum**. In 2010, at the height of the robo-signing crisis, President Obama declined to sign the Interstate Recognition of Notarizations Act, citing consumer protection in mortgages. He did not dismantle electronic mortgage infrastructure. He interrupted one step of its standardization at a moment when the integrity of assignments was in open question. The conflict he was standing in is the conflict that governs personalized supervision:
**Interoperability accelerates institutions. Contestability protects persons.**
### XXV. Assurance and the Rating Agencies: The Manufacture of Ordinal Trust
An observation is not a decision. A model is not an authority. What converts a computation into something an institution may lawfully and defensibly **act upon** is **assurance** — and, above assurance, **rating**.
ICE has disclosed independent assurance over **ICE Data Indices'** implementation of the **IOSCO Principles for Financial Benchmarks**, performed by **PwC**. Precision matters: PwC is not ICE's financial-statement auditor — that is Ernst & Young — and the evidence does not make PwC the corporate auditor of the ING group. Its documented role is **benchmark governance assurance**, which is more consequential than any superficial corporate connection. Assurance does not certify that an index is true. It certifies that the index was produced **within a declared governance and control framework**. Once institutionally trusted, the index governs capital without any downstream user inspecting the world it came from.
Above that sits the layer this argument was missing: the **rating agencies** — Moody's, S&P Global, Fitch, DBRS Morningstar. Their function is to convert an uncertain future into an **accepted class**: a compressed ordinal judgment that thousands of institutions act upon without re-deriving it. A rating is not a fact. It is an **institutionally recognized opinion with the force of a fact**, and the entire architecture of pension mandates, capital requirements, collateral eligibility, and portfolio policy is built to obey it.
Perform the substitution, and the homology is exact.
**A validated risk instrument is a rating agency for persons.**
PATTERN, PCRA, LSI-R, COMPAS, the housing-triage index, the Chicago heat score: each converts a heterogeneous, contested, causally plural human future into a **compressed ordinal class** — minimum, low, medium, high — which courts, wardens, providers, housing authorities, and release authorities act upon **without re-deriving the underlying life**. The instrument is validated, audited, subject to interrater-reliability protocols, reviewed by an Independent Review Committee, and audited biennially by GAO. The First Step Act's apparatus is, structurally, **IOSCO for people**.
And here both systems reveal their shared vulnerability. In 2008 the world learned what happens when an ordinal-trust layer is systematically wrong: everyone had stopped looking at the mortgages, because the rating was doing the looking for them. **The person's tier now performs the same function.** No landlord re-derives the life. No employer re-derives the life. They read the class.
This is why the **managerial translation class** — PwC, Deloitte, EY, KPMG, IBM, Accenture, Infosys — is not decorative in a story about prisons. **A scoring infrastructure becomes scalable precisely when it becomes auditable.** Assurance is not the opposite of power. Assurance is what makes power portable.
### XXVI. The Secondary Market: Why Portability Pays
A loan becomes powerful when it can be **pooled, tranched, rated, serviced, insured, hedged, and sold**. The claim leaves the lender; the obligation stays with the household; everything upstream becomes an asset class.
The correctional economy is rarely described this way, and it should be, because the mechanism is identical: **standardize the event stream, and capital can be structured around it.**
**Payment or default. Foreclosure or modification. Compliance or violation. Attendance or nonattendance. Release success or rearrest.**
Around that stream sits a complete **capital-allocation system** in which predicted or measured human outcomes determine revenue, renewal, procurement, and institutional survival: **per diem payments** for residential reentry beds; **electronic-monitoring fees**, frequently charged to the monitored person; **treatment reimbursements** contingent on documented attendance; **prison telecom commissions** paid back to the agency that selected the vendor; **halfway-house placements** allocated by contract; **grant formulas** conditioned on measured recidivism reduction; **outcome-based procurement**; **social-impact bonds** repaid out of avoided incarceration costs; and **software licenses** for the platforms that generate the metrics on which all of the above depend.
I claim no exchange on which recidivism probabilities trade like mortgage-backed securities; none exists. The claim is duller and stronger: **the institutions surrounding the supervised person have their revenue, renewals, and existence tied to the numbers those persons generate.** Which means the system holds a **structural interest in the persistence of the measured population**. A pay-for-success instrument requires failures to prevent. A monitoring vendor requires people to monitor. A technical-assistance contract requires jurisdictions to assist.
The person is the substrate. The **monetizable event stream** is the product. And once revenue depends on the stream, the incentive to keep it flowing is nobody's malice. It is simply the budget.
### XXVII. Insurance Transmission, Geospatial Coupling, and Proxies
Close the loop between the planetary sensorium and the man counting days. The causal chain runs through **insurance**, the most underrated coercive institution in the developed world. Insurance does not accuse anyone. It merely withdraws. And withdrawal propagates:
**climate model → insurance repricing or withdrawal → mortgage impairment (no policy, no loan) → declining property liquidity → falling assessed values → shrinking municipal revenue → reduced public services, policing pressure, school degradation, program cuts → worsening local conditions → higher individual risk classifications for the people who live there.**
That chain begins in a satellite and ends in a supervision tier. The last step is not metaphorical: neighborhood conditions, housing instability, and employment scarcity are **explicit dynamic criminogenic needs** in the risk instruments. **Planetary sensing returns, six institutions later, as a person-level risk variable.**
This is where **geospatial coupling** becomes the hinge. Climate exposure, mortgage value, insurability, policing intensity, incarceration exposure, school quality, broadband, environmental burden, eviction density, and property liquidity all attach partly to **place**. A person can be down-ranked not for anything they did but because of an **address** — a census tract, a flood zone, a heat map, a redlined legacy. Robert McDaniel's most predictive attribute was his block. Call the intersection what it is: **geospatial meritocracy**.
Which brings the argument to **proxy variables**. The models are formally innocent; protected characteristics are excluded. But **zip code proxies race and class**; **employment gaps proxy illness, disability, and caregiving**; **address instability proxies poverty**; **association variables proxy neighborhood structure**; **arrest density proxies police deployment rather than conduct**; **digital behavior proxies ideology and mental state**. The model reconstructs the prohibited variable from its correlates and then reports, accurately, that it never used it.
**The claim to neutrality rests on the absence of the forbidden variable, not on the absence of its computational substitute.** That is the entire epistemic trick of contemporary meritocracy, in one sentence, and it operates identically on a floodplain, a borrower, a census tract, and a man on supervised release.
## PART SEVEN — THE LAW AND THE ASYMMETRY
### XXVIII. The Legal Permission Structure
The invisible architecture is not built by technology. It is built by **legal asymmetry**.
The **Fair Credit Reporting Act** regulates consumer reports and creates adverse-action and dispute rights — but only for data classified as a consumer report, used for a covered purpose, by a covered furnisher. The **Equal Credit Opportunity Act** and **Fair Housing Act** prohibit discrimination on protected bases, and reach proxies only through disparate-impact doctrine, itself contested and politically mobile. The **Gramm–Leach–Bliley Act** governs financial privacy with broad affiliate-sharing carve-outs. The **Privacy Act** binds federal agencies, with routine-use exceptions wide enough to drive a data-sharing agreement through. **HIPAA** covers a narrow band of clinical actors and almost nothing else that touches a supervised person. State **data-broker registration** laws demand disclosure of existence, not of inference. And the **public-record exception** underwrites everything: once a fact is a public record, its aggregation, resale, and modeling are largely unregulated.
The shape of the asymmetry is now visible. **Data may move lawfully across institutional boundaries, while the subject's right to inspect, correct, contextualize, or appeal the resulting inference remains fragmented, domain-specific, and in the most consequential cases entirely absent.**
You can dispute a credit item. You cannot dispute a risk tier. You can demand adverse-action notice from a lender. You cannot demand it from a mentor's case note, a threat-assessment team, or a triage algorithm that placed you eleventh on a housing list. **The law regulates the channel and ignores the inference** — which is exactly backwards, because the inference is the thing that acts.
### XXIX. Constructive Custody: What the Courts Already Know
The most remarkable thing about the invisible prison is that American law has already conceded its existence — repeatedly, and in the government's favor.
In ***Jones v. Cunningham*** (1963), the Supreme Court held that a person on parole remains **in custody** for habeas purposes, because the conditions of parole significantly restrain freedom. In ***Hensley v. Municipal Court*** (1973), the Court extended this to a person released on his own recognizance, reasoning that he was **subject to restraints not shared by the public generally** — a doctrinal formula of extraordinary generality that does not mention walls. In ***Morrissey v. Brewer*** (1972), the Court held that parole liberty involves significant values within the protection of due process.
Correctional law says the same from the other direction: departments define **administrative custody** as a *status of confinement* rather than a location. A district attorney describing civil commitment put it as plainly as any theorist ever has: **civil commitment is not a place, it's a status.**
So the courts, the corrections departments, and the prosecutors all agree with the thesis of this essay. **Custody is a status.** The disagreement is only about which statuses count.
And that is where the doctrine stops being an ally. Under ***Paul v. Davis*** (1976), reputational harm inflicted by the state is not a constitutional injury unless coupled with the loss of a recognized right or status — the **"stigma plus"** rule. The modern architecture generates stigma constantly: a threat-assessment flag, a risk tier, a heat-list score, a gang-database entry, a watchlist. Whether it also generates the "plus" depends on whether the resulting deprivation — the housing not offered, the job not extended, the program not entered — is characterized as a **right lost** or a **benefit not conferred**.
That characterization is doing more work in American liberty jurisprudence than any doctrine has a right to. And the entire drift of the architecture — from sentence to score, from restriction to non-provisioning, from punishment to *the absence of acceleration* — is a drift **toward the side of the line where due process does not reach**.
The invisible prison is not built where the law is silent. It is built precisely where the law says: *nothing was taken from you.*
### XXX. Pre-Crime: Threat Assessment, Intercept Zero, and the Custom Notification
Prediction has already crossed the line separating punishment from prevention, and it did so under the vocabulary of care.
**Behavioral Threat Assessment and Management**, promoted through the Department of Homeland Security's prevention programs and the Secret Service's threat-assessment research, deploys **multidisciplinary teams** — educators, human-resources staff, clinicians, law enforcement — to identify **persons of concern** exhibiting indicators along a **pathway to violence**, and to intervene during ideation or grievance, *before* any adjudicated offense. Its designers insist, sincerely, that it is **non-punitive**.
Examine the operation rather than the intention. A person is **classified without adjudication**, **discussed across institutions without notice**, **subjected to intervention without conviction**, and **denied any forum in which to contest the classification** — because formally, nothing was done *to* them. They were helped.
That is the Chicago custom notification, generalized and made national.
The **Sequential Intercept Model** performs the same expansion on the clinical side. Originally a map of where a person with behavioral-health needs might be diverted *out* of the justice system, it has grown an **Intercept 0** — crisis services, mobile outreach, co-responder teams — situated *before* any law-enforcement contact at all. Which means the model no longer describes exits from the justice system. It describes **entrances into the supervision system that bypass the justice system entirely**.
This is where the New Penology completes itself. Feeley and Simon's actuarial subject did not need to be guilty. He needed to be **classifiable**. And a classifiable person can be managed, dosed, supported, monitored, and modulated without ever passing through the one institution — the criminal trial — designed to make the state prove something.
### XXXI. The State Does Not Score Itself
Every scored subject in this essay is a person, a household, a property, a firm, or a territory. **No instrument scores the institution.**
The supervised person is assessed on intake, reassessed on a schedule, coded after every contact, monitored continuously, and re-tiered on dynamic factors. Failure carries immediate, individualized, materially consequential penalties.
The institution supervising that person is assessed **episodically**, by a body it does not fear, with findings that carry no automatic consequence. In March 2023 the Government Accountability Office [reported](https://www.gao.gov/assets/gao-23-105139.pdf) that the Bureau of Prisons' risk-and-needs system suffered delayed assessments, incomplete and unreliable data, and programming capacity insufficient to deliver the interventions the assessments prescribed. Its January 2026 follow-up found the timeliness and technology problems persisting. Nobody's liberty was restricted. Nobody's tier changed. No agency was placed on intensive supervision.
**The person is governed by continuous scoring. The system governing the person is protected by episodic review and diffuse accountability.**
This is the constitutional heart of the matter, and it is a specification rather than a flourish. The architecture is **structurally incapable of applying to itself the epistemology it applies to everyone else**. Wrongful records, broken referrals, vendor errors, model drift, program shortages, discriminatory proxies: none is converted into an institutional risk score producing institutional consequence.
If risk-and-needs assessment is the correct way to govern a fallible actor whose future conduct threatens the public, then the Bureau of Prisons should have a PATTERN score. So should the vendor. So should the model. That this proposal reads as absurd, while its mirror image reads as evidence-based practice, is the entire ideology of the system rendered in one asymmetry.
## PART EIGHT — THE CLOSURE
### XXXII. Social-Credit Equivalency and Meritocratic Throttling
It is often said that the United States has no social credit system because it issues no single universal score. This is true, and it is not reassuring. It describes a **more robust architecture**, not a weaker one.
Define a **social-credit equivalency system** as any distributed institutional ecology in which multiple forms of recorded behavior, compliance, reputation, eligibility, and modeled risk are translated into **differential access**. The United States possesses one, assembled from openly documented, individually lawful components: credit histories; criminal records; risk-and-needs classifications; insurance scores; tenant screening; background checks; mortgage and title records; identity resolution; platform reputation and ranking; program eligibility; electronic monitoring; treatment-compliance records; coordinated-entry systems; grant databases; community reports; threat-assessment files; predictive-policing scores; and the ordinary institutional watchfulness of employers, landlords, clinicians, and schools.
The system is **federated rather than unitary**. Its power lies not in a master number but in the cumulative, interoperable capacity of institutions to **up-rank, down-rank, accelerate, dampen, admit, delay, redirect, restrict, or exclude**.
**Dampening**: higher borrowing and insurance costs, delayed transactions, reduced liquidity, foreclosure pressure, intensified supervision, repeated verification, program exclusion, narrower employment and housing pathways, increased reporting, restricted mobility, reputational down-ranking, reduced institutional tolerance — and, decisively, **diminished access to the very resources required to improve the score**.
**Acceleration**: cheaper credit, faster approval, greater liquidity, employment and housing access, lighter supervision, earlier discharge, wider mobility, trusted status, reduced reporting, richer informational and social affordances.
Both operate on properties, territories, institutions, and persons. Both are administered as though they were **neutral responses to merit**.
They are not neutral. **The act of provisioning is itself causal.**
**adverse classification → reduced affordances → declining adaptive capacity → adverse outcome → confirmation of classification**
**favorable classification → expanded affordances → increased adaptive capacity → favorable outcome → confirmation of classification**
A down-ranked property faces insurance withdrawal, reduced lending appetite, deferred maintenance, and declining liquidity — producing exactly the deterioration the model anticipated. A down-ranked person faces more conditions, more observation, and fewer opportunities through which favorable evidence could ever be generated. Robert McDaniel is the boundary case that proves the rule: **the intervention licensed by the prediction produced the predicted event.**
The system appears meritocratic because later outcomes correlate with earlier rankings. But the ranking **helped manufacture the outcome** by controlling the provision of stabilizing resources.
Call it **meritocratic throttling**. The system does not merely discover winners and losers. **It modulates their velocity.**
### XXXIII. The Failure Mode: The Sensor Outlives the Covenant
If this architecture's danger were tyranny, we would know what to fight. Its actual pathology is stranger: **abandonment with the instruments left running.**
The classification engine works. The prescription is issued. **The prescribed thing is not there.**
In April 2025, BOP issued a directive capping most Second Chance Act halfway-house placements at **60 days** — rescinding release dates already granted and stranding people who had earned their way out. In the same month, the Office of Justice Programs terminated roughly **365 grants** initially valued near **\$811 million**, including approximately **\$40 million** of Second Chance Act work. The cuts fell heavily on the **technical-assistance intermediaries** — the translation class that trained providers, bridged referral gaps, and made the polycentric architecture cohere where a person meets a caseworker. The Second Chance Act's authorizations had in any case run only through fiscal year 2023.
Assemble it. The person is assessed. The needs are diagnosed. The dosage is prescribed. The credits are earned. And the bed does not exist, the provider has been defunded, the intermediary has laid off its staff, and the interagency body that was supposed to make housing, labor, health, and education answer to one another has been chartered, revoked, sunset, reconstituted, and revoked again.
**The score persists. The service evaporates.**
The benevolent loop becomes an **extraction loop**: the person continues to generate data and receives diminishing support. Risk remains individualized; institutional failure remains diffuse, collective, and entirely unscored.
This is the general law of the architecture, at every scale in this essay. Councils sunset. Executive orders expire. Philanthropies divest. Grant networks dissolve. Consent regimes are revoked by regulators. Public covenants lapse. Yet the registries remain. The data standards remain. The mortgage records remain. The benchmark methodologies remain. The catastrophe models remain. The master person indexes remain. The consumer files remain. The psychometric literature remains. The cell atlases remain. The scores remain. **The institutional memory encoded in software remains.**
Meanwhile the American smart prison itself was captured commercially, exactly as the pattern predicts. Where Finland built a state-financed rehabilitative interface, the United States built a **private-equity tablet estate**: Securus and ViaPath, together holding roughly **80 percent** of the carceral communications market across some five thousand facilities, had deployed more than **600,000 tablets** by mid-2023, up from 400,000 in 2021. Corrections departments receive **commissions** on the traffic — the agency selecting the vendor holds a direct financial interest in the price the prisoner's family pays. In several states people are billed **per minute** to read messages and use media they already purchased. And when a state changes vendors, the archive does not migrate: incarcerated people in Tennessee, New Jersey, Virginia, and California have lost years of family photographs, video messages, and purchased libraries — one man reporting the loss of six thousand dollars of music and the recorded first words of a grandchild — because a procurement decision was made in a room they will never see. In March 2026 a proposed class action in Tennessee advanced a **Takings Clause** theory.
Same cell terminal. Same rehabilitative vocabulary. One is a public covenant with a contestability layer. The other is an **extraction surface wearing a rehabilitation label**, in which a person's memories are hostage to a contract renewal.
**Measurement without reciprocity** is the signature. It is what the climate ledger does to a floodplain, what the mortgage stack does to a house, what the rating does to a municipality, what the platform does to an electorate, and what PATTERN does to a man counting days.
### XXXIV. The Division of Function
Because adjacency is not evidence, and because a map this size collapses into insinuation unless every organ is assigned a job, here is the division of function this essay asserts — and the limit of what it asserts.
**IMF and the capital-flow literature:** macro-level flow governance; the theory of absorption capacity; the legitimation of the throttle.
**ING and the banking hubs (Warsaw, Katowice):** model development, validation, monitoring, governance — the epistemic factory where transnational risk logic is manufactured.
**ICE and NYSE:** market infrastructure, data, indices, benchmark administration.
**MERS:** identity resolution and portability for mortgage claims.
**ICE Mortgage Technology and Black Knight:** origination, servicing, property intelligence, climate analytics.
**Rating agencies:** the manufacture of **ordinal trust** — uncertain futures converted into accepted classes that institutions obey without re-derivation.
**PwC and the assurance firms:** controls testing, methodology assurance, benchmark governance — what makes an inference safe to rely upon.
**IBM, Accenture, Deloitte, Infosys and the integrators:** business-process engineering; the transfer of managerial ontology into public administration.
**Consumer-reporting agencies and data brokers:** person-level portability; the commercial spine along which adverse memory travels across domains.
**Case-management vendors:** the **single view of the offender**; the digital twin of the supervised citizen.
**Platforms:** the informational-affordance layer — ranking, flagging, redirection; damping and acceleration applied to what a person can encounter and therefore become.
**AOUSC, DOJ, BOP, PCRA, PATTERN:** state-authorized behavioral underwriting; conversion of a human future into a tier.
**Predictive policing (the heat list, the custom notification):** pre-adjudication classification coupled directly to intervention at the front door.
**Technical-assistance intermediaries (CSG Justice Center, CEPP, Carey Group, UCCI, PRA):** doctrine, curricula, validation, and translation of federal strategy into local practice.
**Philanthropy (CZI, The Just Trust, FWD.us):** funding of the **decarceration doctrine** — the demand-side condition for a distributed perimeter — and, separately, of the **measurement science** that could one day make the interior of a person as legible as the exterior of a house.
**Biomedical measurement (CZ Biohub Chicago, Allen Institute, BRAIN, MICrONS):** the instrumentation of tissue and circuit; the eye that has not yet been given a verdict to sell.
**Nonprofits, congregations, mentors, employers, landlords, volunteers:** distributed intervention and **ambient sensing** — morally sincere, informationally compartmentalized.
**And the limit.** No documentary evidence establishes that these institutions jointly designed a program of personalized supervision. No procurement record links CZI, the Biohub Network, or Meta to correctional operations. **I assert neither, and the essay does not need either.** What the record establishes is **structural equivalence and institutional interoperability**: that a single managerial ontology — **observe, resolve identity, construct features, score, rate, assure, allocate, modulate, reassess** — was installed in federal probation by consultants, exported into whole-of-government reentry, universalized across the federal prison population by statute, demonstrated on 87 million people by a political consultancy, and is simultaneously the operating logic of land finance, benchmark administration, catastrophe modeling, sovereign risk, and capital-flow management.
**A distributed, interoperable, risk-based system of differential affordance provisioning can emerge from openly documented institutions without any single actor designing or perceiving the totality.** That is not a weaker claim than leap. It is a far harder one to dismantle, because there is no one to arrest and nothing to repeal.
### XXXV. Freedom as the Capacity to Become Less Predictable
There is an alternative, and part of it was drafted by the woman standing three names away from mine in a directory in January 2020.
**Recommendation CM/Rec(2024)5** requires that prison and probation services use AI and related digital technologies **legitimately and proportionately, and only insofar as they contribute to rehabilitation**. It establishes legality, legal certainty and liability; good governance; **transparency, traceability, and explicability**; the **right to human review of a decision**; and human-centred use. It insists these technologies **assist staff rather than replace them**. And it extends public scrutiny across the entire lifecycle — **design, development, provision, use, and decommissioning** — including for private companies acting on behalf of prison services.
Note what appears there and nowhere in the American architecture. **Decommissioning.** An expiry condition for the model. A moment at which the system must be switched off and accounted for.
But safeguards are not a philosophy. Contestability, expiry, human review, and decommissioning describe what a just system must not do. They do not say what freedom **is** in a world of intelligent systems — and without that, every safeguard is eventually optimized around.
So: **freedom is not the absence of restriction.** In an architecture of differential velocity, an unrestricted person can be throttled to a standstill without a single prohibition being issued. Freedom in this regime is more specific and more demanding — **a protected capacity to generate new evidence about oneself**: to escape a stale classification, to access the resources required for self-revision, to maintain contextual integrity across domains, to refuse compulsory legibility, and to act in ways the model did not anticipate without being punished for the surprise.
Which gives the criterion, and it is the only sentence here I would defend as a principle rather than a finding:
**A rehabilitative system is legitimate only when it increases the subject's capacity to become less predictable to the model.**
A system that merely improves its prediction of the person has become **more intelligent about confinement**. A system that gives the person real resources, real privacy, real mobility, real rights, and real opportunities to **invalidate the prediction** has become **intelligent about freedom**.
By that criterion, dosage probation is legitimate exactly to the extent that the two hundred hours are real, available, and delivered — and illegitimate exactly to the extent that the credits are earned into a void. By that criterion, the Finnish terminal is legitimate and the American tablet is not. By that criterion, a risk score with an expiry is a tool and a risk score without one is a sentence. And by that criterion, the most damning fact about the heat list is not that it watched Robert McDaniel. It is that it left him with **no available action that could have falsified it.**
### XXXVI. What Should Be Said Plainly
Nothing in this essay is hidden. Every consequential claim sits in the Federal Register, the United States Code, GAO audits, OLC opinions, DOJ press releases, SEC filings, FTC complaints, IMF working papers, Supreme Court reports, Council of Europe recommendations, philanthropic press releases, university news bureaus, agency manuals, volunteer handbooks, and the government's own litigation briefs. There is no leak, no whistleblower, no sealed archive. **The record is more damaging than any inference could be, precisely because it was public the entire time.**
It became invisible for three reasons. **Time compartmentalization**: the coordinating bodies were rechartered on a cadence faster than institutional memory. **Polycentric complexity**: the architecture is distributed by design, so each participant sees a humane fragment and none sees the aggregate. And **regime ownership**: no administration intended a hidden system; each intended to *own* the system, and every act of ownership required destroying the previous name. The invisibility is the residue of the ownership fight.
The people building this are not villains. The forensic psychologist writing the Council of Europe's AI safeguards is doing serious moral work. The pediatrician funding the measurement of inflammation is trying to end disease. The philanthropist funding the emptying of prisons is right that the prisons should be emptied. The social worker on Robert McDaniel's porch was there to offer him a job. **Every single one of them is helping. That is the finding.**
So the question the smart prison poses in its purest form is the question the whole architecture poses everywhere at once:
**Will intelligence be used to enlarge the person's world — or to make the perimeter so responsive, so benevolent, and so personally fitted that the person no longer recognizes it as a perimeter?**
The wall is no longer the primary instrument of confinement. The decisive instrument is the **differential velocity of permission**. A person may remain physically present in open society while housing, employment, credit, mobility, communication, reputation, institutional trust, and access to opportunity are continuously accelerated or dampened around them. The prison is not necessarily a place the subject enters. It can be an environment synthesized around the subject from scores, registries, reports, incentives, and conditional affordances.
**Its walls are rendered at runtime.**
---
[**Bryant McGill**](https://bryantmcgill.com/about/) is a Wall Street Journal and USA Today Best-Selling Author, founder of Simple Reminders, and architect of the Polyphonic Cognitive Ecosystem. A Congressionally Recognized Ambassador of Goodwill and United Nations appointed Global Champion, his work spans naval intelligence systems, computational linguistics, and civilizational governance architecture.
---
## References
**Justice, reentry, and the coordinating bodies**
- [Executive Order 13826—Federal Interagency Council on Crime Prevention and Improving Reentry](https://www.federalregister.gov/documents/2018/03/12/2018-05113/federal-interagency-council-on-crime-prevention-and-improving-reentry) — Federal Register, 12 March 2018 (§5 revocation; §6 three-year termination)
- [The Attorney General's Reentry Council](https://nij.ojp.gov/topics/articles/attorney-generals-reentry-council) — National Institute of Justice
- [The Federal Interagency Reentry Council: A Record of Progress and a Roadmap for the Future](https://nationalreentryresourcecenter.org/resources/federal-interagency-reentry-council-record-progress-and-roadmap-future) — National Reentry Resource Center, 2016
- [First Step Act of 2018, Public Law 115–391](https://www.congress.gov/115/plaws/publ391/PLAW-115publ391.pdf) — §101 risk and needs assessment; §505 federal interagency reentry coordination
- [Coordination to Reduce Barriers to Reentry: Report of the Attorney General's Reentry Coordination Council](https://www.justice.gov/archives/opa/press-release/file/1497911/dl) — U.S. Department of Justice, April 2022
- [DHS Contributions to the Alternatives, Rehabilitation, and Reentry Strategic Plan](https://www.dhs.gov/dhs-contributions-alternatives-rehabilitation-and-reentry-strategic-plan) — the Alternatives and Reentry Committee under E.O. 14074
- [Second Chance Act Programs](https://bja.ojp.gov/program/sca-programs/overview) — Bureau of Justice Assistance
- [National Reentry Resource Center](https://nationalreentryresourcecenter.org/) — Council of State Governments Justice Center
**Risk assessment, dosage, and the managerial origin**
- [Federal Probation: A Journal of Correctional Philosophy and Practice, September 2020](https://www.uscourts.gov/sites/default/files/fedprobation-sept2020-508_0.pdf) — PCRA, STARR, motivational interviewing, "control with a working alliance"
- [Evidence-Based Practices: Assisting Reentry to Our Communities](https://www.utd.uscourts.gov/sites/utd/files/EBP_0.pdf) — PCRA dynamic factors; the PwC/IBM strategic assessment of federal probation
- [Dosage Probation: Rethinking the Structure of Probation Sentences](https://www.cepp.com/documents/Dosage%20Probation%20Rethinking%20the%20Structure%20of%20Probation%20Sentences.pdf) — Center for Effective Public Policy / National Institute of Corrections
- [The New Penology: Notes on the Emerging Strategy of Corrections and Its Implications](https://onlinelibrary.wiley.com/doi/10.1111/j.1745-9125.1992.tb01112.x) — Malcolm Feeley and Jonathan Simon, *Criminology*, 1992
- [Bureau of Prisons Should Improve Efforts to Implement its Risk and Needs Assessment System (GAO-23-105139)](https://www.gao.gov/assets/gao-23-105139.pdf) — Government Accountability Office, March 2023
- [First Step Act Earned Time Credits](https://www.ussc.gov/education/first-step-act-earned-time-credits) — United States Sentencing Commission
- [The Sequential Intercept Model](https://www.samhsa.gov/criminal-juvenile-justice/sim-overview) — SAMHSA / Policy Research Associates
**Chicago: the instrumented city**
- [Heat Listed](https://www.theverge.com/c/22444020/chicago-pd-predictive-policing-heat-list) — Matt Stroud, *The Verge*, May 2021; the Robert McDaniel case
- [Chicago Police Department ends use of the Strategic Subject List](https://chicago.suntimes.com/city-hall/2020/1/27/21084030/chicago-police-strategic-subject-list-party-to-violence-inspector-general-joe-ferguson) — *Chicago Sun-Times*, January 2020
- [Strategic Subject List — Historical](https://data.cityofchicago.org/Public-Safety/Strategic-Subject-List-Historical/4aki-r3np) — City of Chicago Data Portal (the released dataset: 398,684 individuals)
- [Predictive Policing: Chicago's Strategic Subject List](https://teamupturn.gitbooks.io/predictive-policing/content/systems/chicago.html) — Upturn
- [Advisory Concerning the Chicago Police Department's Predictive Risk Models](https://igchicago.org/2020/01/23/advisory-concerning-the-chicago-police-departments-predictive-risk-models/) — Chicago Office of Inspector General, January 2020
- [Array of Things](https://arrayofthings.github.io/) — University of Chicago and Argonne National Laboratory urban sensing project
- [The Pretrial Fairness Act](https://www.illinoiscourts.gov/) — Illinois; cash bail abolished effective 18 September 2023
**Chan Zuckerberg Initiative, Meta, and the measurement layer**
- [Chan Zuckerberg Initiative Announces \$450 Million to Accelerate Criminal Justice & Immigration Reform](https://chanzuckerberg.com/newsroom/chan-zuckerberg-initiative-announces-450-million-to-accelerate-criminal-justice-immigration-reform/) — CZI, 27 January 2021 (the divestment of the justice portfolio to the Justice Accelerator Fund / The Just Trust and FWD.us)
- [New Chan Zuckerberg Biohub Chicago Will Engineer Technologies to Measure Human Biology](https://chanzuckerberg.com/newsroom/biohub-chicago-engineer-technologies-measure-human-biology/) — CZI, 2 March 2023 ("embedding thousands of sensors and sampling probes in tissues")
- [CZI, State and City Leaders Launch CZ Biohub Chicago](https://biohub.org/news/czi-state-and-city-leaders-launch-chan-zuckerberg-biohub-chicago/) — CZ Biohub, 5 October 2023
- [Illinois to co-lead new CZ Biohub in Chicago](https://news.illinois.edu/illinois-to-co-lead-new-cz-biohub-in-chicago/) — University of Illinois Urbana-Champaign News Bureau (the "instrumented tissue mimic")
- [The BRAIN Initiative](https://braininitiative.nih.gov/) — National Institutes of Health
- [MICrONS: Machine Intelligence from Cortical Networks](https://www.iarpa.gov/research-programs/microns) — IARPA; the Allen Institute's cubic-millimeter cortical reconstruction
- [FTC Imposes \$5 Billion Penalty and Sweeping New Privacy Restrictions on Facebook](https://www.ftc.gov/news-events/news/press-releases/2019/07/ftc-imposes-5-billion-penalty-sweeping-new-privacy-restrictions-facebook) — Federal Trade Commission, July 2019
- [Revealed: 50 million Facebook profiles harvested for Cambridge Analytica](https://www.theguardian.com/news/2018/mar/17/cambridge-analytica-facebook-influence-us-election) — *The Guardian / The Observer*, March 2018
**Electronic monitoring and its origins**
- [The Wandering Prison: Electronic Monitoring's Rehabilitative Origins](https://scholarship.claremont.edu/) — Robert Gable on the Schwitzgebel brothers' Streetcorner Research
- [Left to Their Own Devices](https://www.themarshallproject.org/) — The Marshall Project, on Judge Jack Love and the Goss-Link
**The pandemic and the dissolving perimeter**
- [Home Confinement Under the CARES Act — Final Rule](https://www.federalregister.gov/documents/2023/04/04/2023-07063/office-of-the-attorney-general-home-confinement-under-the-coronavirus-aid-relief-and-economic) — Federal Register, 4 April 2023
- [Discretion to Continue the Home-Confinement Placements of Federal Prisoners After the COVID-19 Emergency](https://www.justice.gov/olc/file/1457926/dl) — Office of Legal Counsel, 21 December 2021
- [Home Confinement of Federal Prisoners After the COVID-19 Emergency](https://www.justice.gov/olc/file/1355886/dl?inline=) — Office of Legal Counsel, 15 January 2021
**Constructive custody and stigma-plus**
- [Jones v. Cunningham, 371 U.S. 236 (1963)](https://supreme.justia.com/cases/federal/us/371/236/)
- [Hensley v. Municipal Court, 411 U.S. 345 (1973)](https://supreme.justia.com/cases/federal/us/411/345/)
- [Morrissey v. Brewer, 408 U.S. 471 (1972)](https://supreme.justia.com/cases/federal/us/408/471/)
- [Paul v. Davis, 424 U.S. 693 (1976)](https://supreme.justia.com/cases/federal/us/424/693/)
**Capital flow, model governance, and the translation class**
- [Managing Large-Scale Capital Inflows: The Case of the Czech Republic, Poland and Romania (IMF WP/12/138)](https://www.imf.org/external/pubs/ft/wp/2012/wp12138.pdf) — Leonor Keller and Ibrahim Chowdhury, May 2012
- [Working in ING Hubs Poland](https://careers.ing.com/en/Working-in-ING-Hubs-Poland) — Katowice and Warsaw: model development, monitoring, validation, governance
- [Katarzyna Zajdel-Kurowska](https://www.worldbank.org/en/about/people/k/katarzyna-zajdel-kurowska) — World Bank
**Land finance, registries, ratings, and benchmarks**
- [Intercontinental Exchange Completes Acquisition of MERS](https://ir.theice.com/press/news-details/2018/Intercontinental-Exchange-Completes-Acquisition-of-MERS/default.aspx) — ICE, October 2018
- [Intercontinental Exchange Form 10-K, FY2018](https://www.sec.gov/Archives/edgar/data/1571949/000157194919000003/ice2018123110k.htm) — completion of full MERS ownership, 3 October 2018
- [FTC Secures Settlement with ICE and Black Knight](https://www.ftc.gov/news-events/news/press-releases/2023/08/ftc-secures-settlement-ice-black-knight-resolving-antitrust-concerns-mortgage-technology-deal) — Federal Trade Commission, August 2023
- [IOSCO Principles for Financial Benchmarks](https://www.iosco.org/library/pubdocs/pdf/IOSCOPD415.pdf) — International Organization of Securities Commissions
- [Why President Obama is Not Signing H.R. 3808](https://obamawhitehouse.archives.gov/blog/2010/10/07/why-president-obama-not-signing-hr-3808) — the Interstate Recognition of Notarizations Act, October 2010
**Smart prisons and ethical governance**
- [Recommendation CM/Rec(2024)5 regarding the ethical and organisational aspects of the use of artificial intelligence and related digital technologies by prison and probation services](https://rm.coe.int/cm-rec-2024-5/1680b1fec1) — Council of Europe, 9 October 2024
- [AI in prisons and probation: new Council of Europe recommendation](https://www.coe.int/en/web/portal/-/ai-in-prisons-and-probation-new-council-of-europe-recommendation-aims-to-ensure-respect-for-human-rights-and-dignity) — Council of Europe
- [The Implementation of the Council of Europe Recommendations for the Use of AI in Prison and Probation Services](https://www.europris.org/events/ai_webinar/) — EuroPris; Pia Puolakka's role and the Smart Prison chronology
- [Smart prisons](https://www.rikosseuraamus.fi/en/sentenced-for-a-crime/serving-a-prison-sentence/visits-and-contacts-with-the-outside-world/smart-prisons/) — Prison and Probation Service of Finland
- [In Finland, prisoners are being taught crucial AI skills](https://www.wired.com/story/finland-ai-prisons) — WIRED
**Commercial capture of the carceral interface**
- [Prison telecom providers shift strategy by exploiting tablet services](https://prismreports.org/2024/12/09/prison-telcom-providers-exploit-tablet-services/) — Prism
- [Incarcerated People Lose Treasured Media When Prisons Change Tablet Contracts](https://theappeal.org/prison-tablets-contracts-incarcerated-people-lose-treasured-media/) — The Appeal
- [Pay-for-Play Tablets: The Costly New Prison Paradigm](https://www.prisonlegalnews.org/news/2025/mar/1/pay-play-tablets-costly-new-prison-paradigm/) — Prison Legal News
- [Canceled DOJ Grants Threaten Bipartisan Work to Support People Released from Prison](https://www.brennancenter.org/our-work/research-reports/canceled-doj-grants-threaten-bipartisan-work-support-people-released) — Brennan Center for Justice
- [DOJ Funding Update: A Deeper Look at the Cuts](https://counciloncj.org/doj-funding-update-a-deeper-look-at-the-cuts/) — Council on Criminal Justice
**Framing**
- [Climate & Meritocracy: How Public Weather Data Became Private Risk Scores](https://bryantmcgill.blogspot.com/2025/11/climate-meritocracy.html) — Bryant McGill
- [Bryant McGill — Global Justice Resource Center](https://globaljusticerc.org/bryant-mcgill/)
---
#SmartPrisons #GlobalJustice #Prison #prisons #criminology #AI #Chicago #CZI #Biohub #Zuckerberg #CambridgeAnalytica #PredictivePolicing #MERS #RiskScoring #FirstStepAct #SystemsHistory #DistributedCustody #ActuarialGovernance #EO13826 #PATTERN #Schwitzgebel #tether #Y2020 #AnkleBracelet #icu
"In cybernetic systems, ethical considerations arise when the observed becomes aware of the observer. The feedback loop of surveillance changes both parties."– Stafford Beer
ORCID iD: 0009-0003-2345-5390
Posts
0 Comments