Daniel sent us this one, and it builds directly on a conversation we just had about Israel's fractured party system. He's zooming in on the policy machinery itself — or the lack of it. His question is basically: if every citizen with a vote is a consumer of policy, how do you tell whether a party's platform is backed by actual research or just vibes? What does a well-oiled policy research machine even look like, and what are the concrete indicators a voter can use to spot the difference?
This is the question that actually makes the fragmentation problem matter. Because if you can't tell the difference between a party that spent eighteen months stress-testing a housing proposal and a party that drafted a three-page manifesto over a weekend, then the whole system starts rewarding the wrong thing.
That's the erosion Daniel's pointing at. A party can form overnight here — literally file the paperwork, slap a name on it, run on one person's face — and dissolve just as fast. But real policy doesn't work on that timeline. It can't.
And the Israel Democracy Institute has been tracking this for years. The Knesset has a hundred and twenty seats, and over the past two decades the number of factions has regularly exceeded ten, sometimes pushing toward fifteen. A lot of those are what they call "personal parties" — vehicles built around a single politician who broke away from a larger faction, often carrying just two or three seats, sometimes fewer.
The political equivalent of a solo founder incorporating an LLC in their living room and calling it a company.
That's the startup metaphor Daniel used, and it's a good one. These breakaway parties pop up overnight, they're branded around the founder's name or a single grievance, and they often dissolve within one or two election cycles. But here's the thing — the startup metaphor also works in reverse for the established parties.
A well-established party is more like a company that's been around for decades. It has departments. It has institutional memory. It has a research division that functions year-round, not just during campaign season. When Likud or Labor in their heyday wanted to produce a major economic platform, they didn't just have the party leader sit down with an advisor and brainstorm bullet points. They had teams of economists, legal experts, sector specialists — people who'd been working on these issues for years, sometimes across multiple governments.
The question isn't really "does this party have policies" — every party has policies, or at least something that looks like policies on a website. The question is whether there's an actual machine behind them. And that machine is largely invisible to voters.
And that's the core problem we want to dig into. Most voters encounter policy as a finished product — a press release, a campaign ad, maybe a bullet-point list on a party's Facebook page. They never see the six to eighteen months of research, the stakeholder consultations, the cost-benefit analyses, the legal reviews, the multiple drafts that got shredded before anything went public.
Six to eighteen months. That's the actual timeline for a credible policy proposal?
From initiation to public release, yes. And that's for a party that already has the machinery in place. If you're starting from scratch — which is what these breakaway parties are doing — you simply cannot compress that timeline without losing the substance. You end up with what looks like policy but is really just a positioning statement.
Let's define what we're actually talking about when we say "policy research machinery." Because I think most people hear that phrase and picture something vague — a think tank somewhere, maybe some academics on retainer. But it's much more specific than that.
And that's exactly where we should go next.
A proper policy research department is staffed year-round, not assembled six weeks before an election. You've got dedicated economists, legal experts, sector specialists — people whose entire job is to take a problem, study it from multiple angles, and produce something that can survive contact with reality.
Survive contact with reality. That's the phrase. Because a lot of what these startup parties produce wouldn't survive contact with a stiff breeze, let alone an actual Knesset committee.
And the IDI has documented this gap in pretty stark terms. They've analyzed party platforms across multiple election cycles and found that established parties routinely publish fifty to a hundred-plus page policy documents with named authors, cited data, and acknowledged tradeoffs. Meanwhile, some of these breakaway parties submit platforms that are three to ten pages of mostly aspirational language.
Three to ten pages. That's not a policy platform — that's a LinkedIn post with formatting.
It's a mood board. And look, I'm not saying length equals quality. A hundred pages of filler is still filler. But when a party can't even produce enough substance to fill a pamphlet, that tells you something about whether there's any machinery behind the words.
The core difference you're drawing is institutional memory versus personality. The established party has a research pipeline that predates the current leader and will outlast them. The startup party is the leader — their instincts, their grievances, their personal brand. When the leader walks, the party evaporates.
That's exactly why the IDI flags this as a governance stability problem. When you have factions that are essentially one-person vehicles, policy continuity becomes impossible. You can't do multi-year planning on housing or healthcare or infrastructure if the party you're negotiating with might not exist in eighteen months.
Which brings us to the question Daniel's really asking. If I'm a voter, and I'm looking at two parties that both say they'll fix the cost of living — one has a hundred-page document with cost projections and named economists, the other has a slogan and a headshot — how do I actually evaluate that difference? What are the telltale signs?
That's the voter-as-consumer framing Daniel proposed, which I think is genuinely useful. Because right now, most voters consume policy the way they consume advertising — they react to the emotional resonance, the branding, whether it aligns with what they already believe. They're not equipped to ask, "Where did this come from? Who wrote it? What happens when it meets a spreadsheet?
The political equivalent of reading the ingredients label.
And the ingredients label for most of these startup parties would read "vibes, personal ambition, and a graphic designer who knows how to make a PDF look professional.
If we're going to give people an actual toolkit for reading that ingredients label — which is where I think we need to go — we first need to understand what the real machinery looks like when it's working. What are the actual components? Who's in the room? What's happening during those eighteen months?
That's the right next step. Because once you know what a functioning policy apparatus looks like, the absence of one becomes much easier to spot.
Let's walk through what's actually inside one of these machines. A well-resourced party policy department — and I'm thinking of what Likud maintained during its dominant periods, or what Labor had in the nineties — typically runs several specialized units. You've got an economics desk, a legal affairs desk, a social policy desk covering health, education, welfare. Each of those has dedicated staff — not volunteers, not campaign-season consultants, but people on payroll whose job description is "produce policy research.
These aren't the people you see on television. They're not the faces of the party.
They're invisible. And that's by design, partly — the party leader and Knesset members are the public-facing layer. But the research staff are the ones reading the academic literature, running the numbers, consulting with the relevant ministries to understand what's actually administratively feasible. They're the ones who know that a housing proposal isn't just about zoning — it's about infrastructure capacity, municipal budgets, construction sector labor supply, interest rate sensitivity. They map the whole system.
This runs year-round, not just when an election's called.
That's the critical piece. Because if you only start doing policy work six weeks before voters go to the polls, you can't do any of that mapping. You're guessing. You're writing what sounds good. The IDI has actually looked at this — they've found that parties with permanent research staff produce platforms that reference specific government data, name the studies they're drawing from, and acknowledge implementation risks. Parties without that infrastructure produce platforms that are essentially collections of desired outcomes with no connective tissue.
"we will lower housing prices" versus "here is the specific regulatory change, here is the projected cost to the treasury, here is the timeline, here's who we consulted, and here's what could go wrong.
That's the spectrum. And the eighteen-month timeline I mentioned earlier — that's not arbitrary. A typical major policy proposal goes through distinct phases. Months one through four are research and evidence gathering — commissioning studies, reviewing what other countries have tried, interviewing stakeholders. Months five through eight are drafting and internal stress-testing — the economists tear apart the legal team's assumptions, the legal team flags constitutional issues, everyone argues about the cost projections. Months nine through fourteen are external consultation — bringing in industry groups, municipal representatives, sometimes public hearings. Months fifteen through eighteen are revision, leadership review, and preparing the public-facing materials.
Eighteen months of people in rooms arguing about spreadsheets. That's the unsexy reality behind a policy that actually works.
Here's the contrast that makes the startup party problem so stark. A breakaway faction with three Knesset seats, formed six weeks before an election deadline, has none of this. They have the leader, maybe one aide, and a deadline. Their platform gets written in days or weeks — a series of positions that sound plausible but haven't been tested against administrative reality, legal constraints, or fiscal arithmetic.
The voter sees both — the eighteen-month product and the weekend product — presented in the same format. Similar fonts, similar rhetoric, similar promises. There's no label that says "this one was stress-tested" and "this one was written on a Thursday.
Which is exactly why the IDI's platform analysis work is so valuable, and so underused. They've developed criteria for evaluating policy depth — does the platform cite specific data sources? Does it name the experts or institutions consulted? Does it acknowledge tradeoffs and costs, or is it all benefit with no price tag? Does the party's position show evidence of having evolved over time as new information emerged, or is it frozen in place?
That last one is interesting — evolution as a sign of seriousness, not weakness.
Because a party that actually does research learns things. The economic data changes. A pilot program in another country fails. A stakeholder consultation reveals an obstacle nobody anticipated. A serious party updates its positions accordingly. A personality-driven party can't do that without undermining the leader's brand as someone who just knows the answers.
Rigidity becomes a tell. If a party's housing platform in twenty twenty-four is word-for-word identical to what it was in twenty twenty, either they achieved prophetic perfection or nobody's been doing any work.
We both know which one's more likely. I want to give a concrete example of what the real machinery produced, because it makes the contrast tangible. In the lead-up to the twenty twenty-two election, Likud's economic platform — whatever you think of its conclusions — was developed over roughly eighteen months with input from external economists, including people from the Bank of Israel's research department and several academic institutions. The document ran over eighty pages, included specific fiscal projections, named the modeling assumptions, and identified sectors where the proposed reforms would create transitional costs.
On the other side of the spectrum?
I looked at one breakaway party from that same cycle — I won't name them, it's not about shaming individuals — whose entire economic platform was four pages. Three of those pages were a critique of the government's record. The actual policy proposals amounted to seven bullet points, none of which included a cost estimate, a timeline, or an identified funding source.
So the ratio here is roughly eighty to four. Twenty to one. And both of these were presented to voters as "economic plans.
That's the market failure Daniel's question points toward. When the consumer can't distinguish the eighty-page stress-tested document from the four-page wish list, the incentive to invest in the eighty-page document collapses. Why spend eighteen months and serious money on research if it doesn't move votes any more effectively than a well-designed flyer?
Which means the entire political marketplace starts optimizing for the wrong thing. Not better policy — better branding.
That's where we need to go next. Because if we understand what the machinery looks like, the question becomes how voters can actually spot its presence or absence from the outside.
The toolkit question. What are the concrete tells?
Let's build it. Five things a voter can actually look for, right now, without needing a policy degree. First one: does the proposal name specific costs or tradeoffs?
That's the single biggest tell. A serious policy document will tell you what it costs, who pays, and what gets worse in order for something else to get better. A vibes-based platform is all upside — "we will improve healthcare and lower taxes and increase defense spending" with no acknowledgment that these things pull in opposite directions.
The magic spreadsheet where every variable improves simultaneously.
Second indicator: is there a named author or research team? Not the politician who announced it — the people who wrote it. If a party can't or won't tell you who did the actual analytical work, assume nobody did.
That's not just about accountability. It's about whether the party even has people who could be named. If the answer is "the leader's vision," that's not a research team.
Third: does the document reference existing data or studies? Not "experts say" or "studies show" — actual citations you could look up. The IDI's platform analysis specifically flags this. A research-backed proposal situates itself in what's already known. A vibes proposal just announces conclusions.
Fourth — and I liked this one when you raised it earlier — has the position changed over time? Does the party's housing platform from three years ago look different from today's, or is it copy-pasted? Evolution signals that someone's been paying attention to new data.
Fifth: what's the party's track record on implementation? Not their promises — their delivery. Did they propose things in previous cycles and actually fight for them, or did the platform evaporate the day after the election? A party with institutional memory can point to bills they advanced, committee work they did, amendments they secured even from opposition.
The five-point checklist: costs and tradeoffs, named authors, cited data, position evolution, implementation track record. That's the ingredients label.
Here's the transparency test that ties it together. A party that's serious about policy development is usually open about its process — they publish consultation submissions, they release research briefs, they hold events where the policy team presents findings before leadership turns them into talking points. Parties that don't do this are filtering everything through a spokesperson, and that filtering is itself a signal.
Yesh Atid in twenty thirteen is actually the interesting counterexample here. New party, no institutional history — but Lapid invested in a proper research operation before the campaign. They produced detailed sector-by-sector proposals with named economists attached. It wasn't a hundred-page legacy document, but it had substance you could audit.
That proves it's not just about age — it's about the decision to invest in the machinery. A new party can choose to do the work. Most don't, because the electoral payoff for doing it is uncertain. Which brings us to the shelf-life problem.
The policy that sounds good but hasn't been kicked around by anyone who knows how a ministry actually functions.
Startup parties are especially vulnerable to this because they have no institutional memory of what's been tried before. They propose things that were attempted and failed fifteen years ago, or that sound elegant until you map them onto the administrative reality of six different ministries with overlapping jurisdictions. Established parties have people who remember why the last attempt collapsed — but the flip side is they can become so aware of obstacles that they stop proposing anything ambitious.
You get ossification on one side and naivete on the other. Neither produces good policy.
The knock-on effect is where this really bites. When voters can't distinguish the stress-tested proposal from the wish list, the entire political marketplace degrades. Parties stop competing on policy quality because policy quality doesn't win elections. They compete on charisma, on outrage, on how effectively they can dominate a news cycle. The fragmentation trend we've been talking about is a symptom of this — if a solo operator with a grievance and a social media presence can get three seats without ever producing a serious policy document, why would anyone invest in the machinery?
The market's sending a signal, and the signal is: don't bother. Branding beats research. Vibes beat evidence. And every cycle that this holds true, the incentive to build real policy capacity erodes further.
Which means the toolkit we just laid out isn't just about being an informed voter. It's about changing what the market rewards. If enough people start asking these five questions — and more importantly, sharing the answers — parties notice. They hire researchers instead of just PR people.
The consumer has more power here than they think. But only if they know what to look for.
Let's make this usable. Three questions you can ask about any party before you vote — and I mean literally, pull up their website and ask these. First: can you find the actual policy document, not just the press release announcing it?
That's the door-check. If all you can find is a news item saying "Party X unveils comprehensive economic plan" but the plan itself isn't downloadable, or it's behind a registration wall, or it's a two-page summary with no further detail — that's your answer right there.
Second question: does the document acknowledge costs, tradeoffs, or implementation challenges anywhere? Not in a footnote — in the body, as part of the argument.
This one is surprisingly binary. The IDI's analysis found that serious platforms almost always include a section on fiscal implications and risk factors. The vibes platforms don't. They just don't. So if you scan the document and see no mention of budget impact, no admission that anything might go wrong, no acknowledgment that other priorities might compete — that's not confidence, that's avoidance.
Third question: has the party's position on this issue evolved based on new evidence, or is it frozen in amber? You don't need a research team to check this — just look at what they were saying three or five years ago. If the language, the proposals, even the statistics are identical, nobody's been doing fresh work.
I want to add something practical here, because I know this sounds like homework. It's not. We're talking about a fifteen-minute policy audit. Go to the party's website. Look for anything labeled "research," "publications," "policy papers." See if there are named experts attached. See if they've published anything beyond campaign materials — position papers from between elections, responses to government consultations, legislative analyses.
If the "publications" section is just a feed of the leader's Facebook posts, that's a data point.
It's the data point. Because a party that does real policy work produces material when there's no election. They respond to ministry consultations. They publish research briefs. They hold events where the policy team presents findings. If the only thing a party has ever published is campaign ads and press releases, there's no machine there.
Here's why this matters beyond any single election. The health of a democracy depends partly on voters being able to distinguish policy substance from political branding. When the machinery of policy research is invisible — when an eighty-page stress-tested document and a four-page wish list look the same to the average voter — the system has no way to reward the party that did the work.
That's the market degradation we were circling earlier, and it's the thing I find worrying. If voters can't tell the difference, parties stop investing in the difference. The researchers get laid off. The policy desks get hollowed out. And over time, the entire political class loses the capacity to do serious policy work even if they wanted to.
It becomes a skill that atrophies across the system. Not just in one party — across all of them. Because why would you maintain a capability that nobody rewards?
And then you get a situation where even the "serious" parties are running on increasingly thin platforms, because the competitive pressure to invest in depth has evaporated. The fragmentation trend accelerates. More solo operators. More personal parties. More policy documents that are really just branding exercises with better fonts.
The fifteen-minute audit isn't just about being a conscientious voter. It's about being a participant in a market that's currently broken. Every time someone checks whether a party's platform has named authors or cited data or an acknowledgment of tradeoffs, they're casting a small vote for the idea that this stuff should matter.
Parties do notice. They're exquisitely sensitive to what voters actually respond to. If the only feedback they get is social media engagement and polling on slogans, they'll optimize for slogans. If they start hearing — from focus groups, from town halls, from the questions journalists ask — that voters are checking their research credentials, they'll invest in research credentials.
The consumer has leverage. But leverage only works if you use it.
This raises a bigger question that we're going to be grappling with for years to come. If policy research is invisible to voters — and right now it mostly is — how do we actually create incentives for parties to invest in it? Is this a transparency regulation problem, or does it require a cultural shift in how voters evaluate parties?
I think it's both, but the cultural piece has to come first. You can mandate that parties publish their research — and some countries have moved in that direction with varying success — but if voters don't know how to read what's published, the regulation just produces more documents nobody opens.
The compliance exercise problem. Parties file the paperwork, check the box, and the public's relationship to policy depth doesn't change at all.
So the cultural shift has to be about equipping voters with exactly the kind of audit framework we just laid out. Not in a "citizens should do homework" scolding way, but in a "here's how you spot when someone's selling you an empty box" way.
There's a new urgency to this that I don't think has fully landed yet. As AI gets better at generating polished-sounding policy documents — and it's getting better fast — the gap between "sounds good" and "is good" is going to widen dramatically.
This is the part that keeps me up. Right now, a four-page vibes platform still requires a human being to sit down and write it. It might be thin, but there's some floor. With what GPT-five-point-six and its successors can produce, you can generate an eighty-page document with citations, cost projections, and professional formatting in an afternoon. It'll look indistinguishable from the real thing.
It'll be hollow. The citations might not check out. The cost projections might be hallucinated. But by the time anyone figures that out, the election's over.
The AI doesn't stress-test. It doesn't consult stakeholders. It doesn't know that a particular zoning reform was tried in Haifa in two thousand nine and collapsed because of a jurisdictional dispute between three ministries. It just produces text that reads like it knows those things.
The five indicators we talked about — named authors, cited data, position evolution, all of it — those become even more critical. Because the surface-level signal of "this looks professional" is about to become completely unreliable.
That's the forward-looking hook here. The ability to audit policy depth isn't just a nice civic skill. It's going to be essential literacy in a world where anyone can generate a convincing policy document with a few prompts.
Which means we should probably revisit this in a year and see whether any parties have actually started publishing their research processes transparently enough to pass the audit. And whether voters are asking for it.
I'd love to hear from listeners who try the fifteen-minute audit on parties they're considering. Which ones passed? Which ones didn't? What did you find when you actually went looking for named authors and cited data?
Email the show at show at my weird prompts dot com. We're curious what you turn up.
Now: Hilbert's daily fun fact.
Hilbert: In nineteen thirty-seven, a Portuguese engineer on the Azores island of São Miguel built a mechanical tide-predicting computer using brass gears and a hand crank. It accurately forecast local tides for decades but was lost when the workshop was converted to a fish processing plant in the nineteen fifties. No photographs survive.
A tide computer. In the Azores. Turned into a fish plant.
I have so many questions, and I know none of them will ever be answered.
This has been My Weird Prompts. If you enjoyed this, leave us a review wherever you listen — it helps. We're back next week.
