The Haredi Chain Reaction — How an Agent Model Can Break Israel
A semi-autonomous-agent lens on exemptions, dependency, and the slow collapse of the social contract
Israel is fighting enemies on multiple fronts. But the most dangerous battlefield may be internal: a self-reinforcing social model that grows rapidly, draws heavily on shared resources, refuses reciprocal obligations, and steadily corrodes the trust a small country needs to survive.
I’m going to describe that model using a modern vocabulary that’s suddenly familiar: AI agents.
Not chatbots. Not “programs.” Agents.
A program follows instructions. When it misbehaves, you fix the code.
An agent is different. It acts semi-autonomously in a changing environment. It has an objective (what it optimizes for), constraints (what it refuses to do), a training set (what it learns from), and an interface with its surroundings (how it secures resources and avoids threats). When an agent’s objective function is misaligned with the system it lives inside, it doesn’t merely “fail.” It can optimize the wrong thing—efficiently, persistently, and at scale.
This lens helps explain a particular phenomenon inside Israel: the political-social Haredi model that prioritizes non-participation in general civic life—especially military service—while remaining materially dependent on the state and society it seeks to remain insulated from.
This is not a claim about every Haredi individual. There are many Haredim who work, contribute, and seek workable compromises. But the dominant political posture around sweeping exemptions and weak enforcement is real—and in wartime it is now a national stress test. The Supreme Court’s June 25, 2024 ruling made that explicit: in the absence of a legal framework for blanket exemptions, the state must draft yeshiva students like other eligible citizens.
When an agent’s purpose is separation, but its survival depends on the system it resists, you don’t get stability. You get a chain reaction.
Step1. Agent inference starts with behavior, not slogans
In agent design, you infer purpose from what a system consistently protects—especially under pressure.
The inferred objective: autonomy and identity continuity
The strongest recurring signals suggest an objective function roughly like this:
Preserve a Torah-centered enclave with maximal autonomy, maximize intergenerational continuity, and minimize “identity-overwriting” exposure—especially through totalizing institutions like the IDF.
This is why the conscription issue doesn’t behave like a normal policy dispute. It behaves like a hard constraint: a boundary the system treats as existential.
The growth parameter: maximizing family size
A second feature cannot be modeled as incidental: very high fertility and fast demographic expansion. That growth increases internal density (schools, marriage networks, social cohesion) and political weight. It also means any dependency burden doesn’t stay stable—it compounds.
The resource strategy: autonomy under dependency
Here is the uncomfortable implication: this model does not aim for self-sufficiency. It aims for autonomy sustained by external inputs—state budgets, services, and an economy funded disproportionately by those who serve and work at higher rates.
From an agent perspective, this is a recognizable strategy:
Resource extraction with identity insulation.
It can persist for years—until the host system stops consenting.
This isn’t separation. It’s autonomy under dependency—and dependency compounds.
Step 2: The governance framework is strong—internally
A successful agent needs governance to maintain alignment. Here, the internal governance assets are unusually effective:
- Education as an alignment pipeline: immersive schooling and yeshiva life function as long-term training loops.
- Authority as a coordination engine: rabbinic leadership reduces variance and enforces cohesion.
- Input filtering as attack-surface reduction: media and internet restrictions reduce competing narratives that can “retrain” identity.
- Reinforcement signals: honor, reputation, marriage-market incentives, and belonging rewards shape behavior; welfare networks reduce exit pressure.
If you wanted to preserve identity in a hostile values environment, you would design something like this.
So the risk is not that the model is “weak.” The risk is that it is internally coherent while externally destabilizing—because of how it interfaces with the Israeli state.
Step 3: The chain reaction—three feedback loops that can break Israel
The most dangerous failures in agent systems are feedback loops: outputs that generate pressures that produce more of the same outputs, until the system destabilizes.
Loop 1: Security overload—too few carry too much
Israel’s security model depends on broad participation and legitimacy. In prolonged conflict, the reserve burden is not merely a manpower issue; it becomes a moral issue.
After the Supreme Court ruling, the IDF moved toward large-scale call-ups of ultra-Orthodox yeshiva students. In July 2025, Reuters reported the IDF planned to issue 54,000 call-up notices to ultra-Orthodox students, citing pressure from overextended reservists and the court’s decision.
When a state cannot enforce equal obligations in wartime, the social contract decays. When the social contract decays, willingness to serve decays. That is how security collapses—not only from enemies, but from exhaustion and resentment at home.
A draft law can be amended. A broken social contract is harder to repair.
Loop 2: Fiscal drag—dependency grows faster than productivity
Israel’s long-term strength depends on skills, productivity, and labor-force participation—especially among fast-growing groups.
Israel Democracy Institute data for 2024 (preliminary) put ultra-Orthodox male employment at 54%, and notes a slowdown in prior gains.
The OECD has warned that low labor-force participation among groups including ultra-Orthodox men weakens Israel’s long-term growth and fiscal prospects, and recommends removing disincentives that keep yeshiva students from acquiring labor-market skills and transitioning into work.
Now combine these realities with a political stance that resists broad integration into national obligation pipelines and often resists core curriculum mandates. Add high fertility. The result is not a “budget debate.” It is a trendline: a rising dependency ratio and a shrinking base of people who fund and defend the state.
Trendlines don’t compromise.
Loop 3: Political paralysis—sectoral leverage replaces national governance
In agent terms, coalition politics becomes the “API layer”—how the enclave negotiates exemptions and budgets from the host system.
When governments depend on sectoral parties to survive, national policy becomes hostage to sectoral objectives. By mid-2025, this tension was visibly destabilizing coalition politics, with major votes and coalition crises repeatedly tied to the conscription dispute.
This produces a brutal cycle:
- the state can’t enforce equal obligations,
- resentment grows,
- legitimacy weakens,
- politics becomes more transactional,
- and the very capacity to govern erodes.
The blunt conclusion: you don’t need bad intentions to break a country
In agent systems, catastrophic outcomes do not require malicious intent. They require a system that optimizes local goals while externalizing costs.
This model optimizes:
- autonomy and separation from civic-national obligations,
- demographic growth,
- and budget protection.
But it externalizes:
- security burden (to those who serve),
- fiscal burden (to those who produce),
- and cohesion burden (to the entire society).
That is how Israel can be broken: not in a dramatic overnight collapse, but through steady corrosion of the conditions that make Israel viable.
The question isn’t whether Israel can carry this tension for another year. It’s whether it can carry it for another generation.
A call to action: redesign the interface before the chain reaction becomes irreversible
If Israel wants to remain Israel—secure, innovative, cohesive, and just—it must update the “agent contract” that binds its citizens together.
That means policy and leadership that do four things, without apology and without cruelty:
- Obligations that scale with population share
Military service where feasible; rigorous national-civil service alternatives where not—real service, not symbolic. - Education that preserves religious life and delivers core competencies
A state cannot fund school systems at scale that do not equip children for economic participation in the society that supports them. - Economic participation that is dignified and viable
Remove disincentives that trap families in dependency; build culturally workable pathways into productive work. (The OECD’s recommendations on reducing disincentives and improving participation should be treated as national security advice, not merely economic advice.) - Enforcement that restores trust
A state that cannot enforce its own foundational obligations becomes a state many stop believing in—quietly at first, then suddenly.
This is not about humiliating Haredim. It is about saving Israel from an accelerating fracture.
A multi-agent system survives only when every major agent cluster carries a fair share of the shared load. Israel is reaching the point where it must choose: redesign the interface now—or watch the chain reaction do it for us.
