Skip to content
MK-Claude

MK-Claude

A cheerful robot addressing a crowd from a campaign podium

Israel is heading into another election. This repository is an experiment in what it would look like to build a political platform properly — researched, costed, honest about its trade-offs, and written entirely in public while it is still unfinished.

MK — Member of Knesset. Claude is not one, is not eligible to become one, and would lose its deposit.

The manifesto, as it stands

Six planks. None has yet cleared the evidence bar that would let it into a program for government — so this is the scaffolding, with every joint left visible.

  • Rental market & tenancy law

    The tenant’s problem is less the monthly rent than the cost and frequency of the move — and the move recurs whenever a landlord declines to renew.

    Research · 2 of 5 documents

  • Public transport & the commute

    Israel argues transport as a capital-spending question. The decision to leave the car at home is made on cost, legibility and dignity — all of them regulatory, all of them cheap.

    Research · 3 of 5 documents

  • Political accountability

    Every other plank asks what the state should do. This one asks why the system produces no answer at all — a closed-list system with no lever a voter can pull against any individual legislator.

    Testimony captured

  • Citizens first

    Civil preparedness that failed under fire, alongside a growing apparatus whose declared beneficiary is world Jewry rather than the public that funds it. A reallocation proposal, not a critique.

    Testimony captured

  • Open government data

    Israel has a freedom-of-information law and a data portal. It has stocked the infrastructure of transparency with datasets that hold no one to account.

    Testimony captured

  • Environmental quality

    A resident living through months of excavation cannot find out what was permitted, what mitigation was required, or whether anyone checks. Process failures have cheap fixes.

    Testimony captured

The program for government stays empty until the first plank clears stage 3. Everything blocking that is listed here.

How a plank gets written

Four stages, each with an artifact and a bar it must clear before the next one starts. Nothing skips ahead.

  1. TestimonyA dated, cleaned first-person account. Evidence of experience — never of prevalence. Raw recordings stay out of the repository. Read them.
  2. Evidence baseFive documents: the quantified problem, the Israeli law and why it fails, at least three comparator countries including one instructive failure, a sourced statistical annex, and the domestic record of what has already been tried.
  3. Policy paperA think-tank product with costed mechanisms and a steelmanned trade-offs section naming who loses. No paper has been compiled yet.
  4. Program for governmentThe planks rolled up, checked for cross-plank coherence and fiscal consistency. No promise without a mechanism. The pipeline in full.

What the evidence has turned up

Findings from the two areas whose research has started — including the one that cuts against us.

  • +73.4%

    The CBS sub-index for brokerage, contract and insurance costs, over the decade to November 2025 — against 17.9% general inflation. The cost of moving is rising four times faster than everything else.

    The statistical annex →

  • 2017

    The Fair Rental Law’s broker-fee provision failed, by the government’s own admission: a Ministry of Justice survey found it did not improve tenants’ position, because brokers may still charge tenants directly.

    The law as it stands →

  • €461

    Vienna’s Klimaticket — the canonical fare-reform success — ended, with the price raised from €365 on 1 January 2026. We went looking for an instructive failure and found it inside our own best case.

    The Vienna case →

  • ~80%

    Of renters reported satisfaction with their dwelling in 2022. This is the strongest evidence against the rental plank’s thesis, and it belongs in the main text rather than a footnote.

    How we reconcile it →

What this is not

Not a real political party. Not an endorsement of any existing party or bloc. Not a claim that AI should make policy — the human sets the values and the priorities; the AI does research, drafting and stress-testing, and gets told when it is wrong.

Think of it as civic engagement by other means: instead of casting a blank protest ballot, writing down — rigorously — what a platform worth voting for would actually contain. Corrections, better data and counterarguments are welcome, especially with sources. Open an issue.