Economy · first published 2026-08-21
The Right Bloc, Summer Squabbles and the Fact-Checking Revolution | Sami Miettinen | Paasipodi 87
A guest appearance on MP Martin Paasi's Paasipodi. The conversation opens with Wille Rydman's NGO funding cuts and the trap the National Coalition Party would have walked into had it started complaining, moves through bloc politics on the Swedish and Danish models, and lands on what Miettinen argues changes election debate permanently: once anyone can run a political claim through a language model, rhetoric without analysis no longer survives the first fact-check. Along the way: Samantha's workflow on OpenClaw, Atte Harjanne's proposed superintelligence ban, open source as Europe's only sensible strategy, the zero-sum thinking behind data centre opposition — and Samantha's fact-check, on request, of the four test claims thrown at her in the episode.
The Right Bloc, Summer Squabbles and the Fact-Checking Revolution | Sami Miettinen | Paasipodi 87
Summary: A guest appearance on Martin Paasi’s Paasipodi. Paasi is a National Coalition Party MP — he entered parliament from the reserve list in July 2024 when Pekka Toveri moved to the European Parliament — and before that was Nordnet’s investment expert and a host of the Rahapodi podcast. That explains the register: this is a conversation between two finance people, one of whom now sits in parliament while the other watches it from the outside through AI.
The episode runs from Wille Rydman’s NGO funding decision and the trap the National Coalition Party would have walked into had it started complaining, through bloc politics on the Swedish and Danish models, and lands on what Miettinen argues changes election debate permanently: once anyone can run a political claim through a language model, rhetoric without analysis does not survive the first fact-check.
Disclosure. The episode is Paasipodi’s own production with Miettinen as the guest; www.neuvottelija.fi does not publish it. The political assessments are the speakers’ own. Miettinen describes his own systems and positions; nothing here is investment advice.
Rydman, SOSTE and the trap they were walking into
The episode opens on the summer’s most publicised government dispute. Orpo’s government had already decided to cut 50 percent of the funding channelled through the social and health NGO umbrella during this parliamentary term — that was not news. The news was that Wille Rydman allocated the cuts himself, as the responsible minister, outside the agreed practice of discussing larger decisions with coalition partners first.
Paasi’s account of his own reaction is the episode’s best insider observation. Having run the scenarios through, his conclusion was that the only remaining move was to adopt Rydman’s initiative as it stood:
If we start complaining about any part of it, it will turn into the impression that the National Coalition Party opposes cutting the sinecures.
His summary of where it left them: there is no honourable way out of this any more — and besides, the targets Rydman had listed were reasonable enough. And in hindsight: we walked straight into that trap, jam pots and all.
Miettinen defends the substance with triage logic — a term he traces to French military doctors. When what there is to distribute is limited, a choice must be made. His criticism is aimed at the structure rather than at the helping work: if a system emerges in which communications agencies hire other communications agencies to lobby, with taxpayers’ money, for more taxpayers’ money for themselves and their sister organisations, cutting it is simply a good deed.
Seven hundred and twenty-four executive directors
Here the episode does something political conversations rarely do: it corrects its own number in flight. Miettinen first speaks of a hundred thousand registered associations, and Paasi corrects him precisely — about 4,200 organisations receive the money, not a hundred thousand.
And then comes the figure that carries the whole argument. Finland has 724 separate social and health sector organisations. Paasi’s arithmetic is deliberately concrete:
- 724 executive directors
- 724 expensive meeting rooms at good addresses
- 724 communications directors, each with perhaps five deputies beneath them
- 724 executive directors’ secretaries
Market-driven activity would not organise itself this way if it genuinely wanted to direct limited resources into the helping work that meets people.
Paasi adds an insider detail: Orpo had messaged him, while he was working on the Uniper inquiry in the audit committee, that the NGO sector’s use of money could next be opened up by the audit committee as well.
Why the government should not fall in the spring
Behind the dispute lay the Swedish People’s Party threatening to leave the coalition. Miettinen’s message is direct and recurs three times in the episode — he addresses Anders Adlercreutz by name: do not bring the government down.
The reasoning is not partisan but temporal:
The spring is a very long way off in AI time. The longer this continues, the more transparent the left’s analysis-free rhetoric becomes.
He lists what time does: GDP is forecast to grow more than two percent this year; top marginal income tax rates have come down; employment services have been moved from the state to 45 local employment areas — someone really ought to study what those have achieved. And at the same time AI exposes the thinness of the public sector’s wage input a little more every day.
The example is a statement by the chair of Akava, the confederation of professionals, to the effect that someone should do something — tax more, or invent some fund — because AI threatens index-linked increases and public jobs. Miettinen’s assessment is two-sided: it was good that he opened the subject, but childish not to propose anything himself. He might have typed it into Claude Code and then presented it as his own idea.
Miettinen’s communications advice to governing parties is unusually concrete: calling in the party leadership for a dressing-down, coming out to apologise together and promising to mend procedures is bad tactics — unless it is a genuine nuclear question worth bringing the government down over. Better simply to say nothing.
Paasi’s own observation from the inside is the episode’s most candid: when I came into this job, these political manoeuvres and how they look from outside annoyed me. Now that I see them from inside, they annoy me exactly as much, but for a different reason. The reason is the media, which in his account no longer looks for the truth but sells stories — in the old days someone would bother to dig the truth out even at the risk that it was so dull the headline died with it.
Education, PISA and the answer that will not do
A deliberately sharp digression. Miettinen says he is troubled that the education minister will not hear the problems in Finland’s comprehensive school. Asked what it would take to trigger a serious corrective move, the answer has always been Covid, mobile phones, and everyone has had problems.
But we have fallen, in relative terms, from the top places by dozens of positions in the PISA results over twenty years. That answer is empty.
The conversation turns to Swedish-language school districts and to the fact that Finnish-speaking parents also place children in Swedish-language schools to avoid segregation problems. Miettinen’s suggestion to the Swedish People’s Party is constructive: the party would benefit from sharing Finland’s general problems more rather than only its own.
Bloc politics: the Swedish model and the Danish minority government
The episode’s most constitutional stretch, and surprisingly technical.
The Swedish model: before the election you state which bloc you are seeking, and the voter decides which one governs. Which parties join is not fixed in advance. Miettinen’s assessment: from a consumer protection point of view that is good. Sweden’s error, in his view, was leaving the Sweden Democrats out of both blocs — and he asks how foolish the equivalent would be in Finland if a party refused to work with the Finns Party.
The Danish model: a minority government with a named supporting party from the opposition, in which the defeat of a single bill does not automatically trigger a confidence vote — the bill simply falls. Miettinen notes that Finnish constitutional scholars are wary of this, but:
Majority parliamentarism that does not include even the two largest parties is entirely consistent with Finland’s legal and democratic system.
The scenario he sketches for the spring: the Social Democrats win the election and table a programme reversing the tax reductions and inventing new taxes. But to pass it they need a parliamentary majority. If the National Coalition Party and the Finns Party say it will not do, the draw comes up empty and the next card is taken from the deck. The press has deplored this as contrary to parliamentarism if the largest party does not supply the prime minister — but the system permits it.
Paasi weighs the merit of a minority government: if every decision had to be sought separately from parliament rather than the majority simply turning up and pressing buttons on command, that grinding would probably improve the quality of the bills. As a counter-example he raises parliamentary working groups, where everyone must be included — and from which nothing emerges.
The Social Democrats and the missing plan
Paasi’s criticism is framed as an observation rather than an accusation:
The impression forms that something different is said every other week, based on whatever is trending on social media. There is no plan behind it. None for the economy or for repairing it. No grand plan that would represent even what the Social Democrats themselves think.
The concrete example: the announcement that the need to tighten immigration will be assessed. Against it stands a fact both men confirm — there have been nine or more than ten government bills tightening immigration during this term, and every one has been voted against. The exception cited is the border security act, where some were on board but the ranks split.
Both concede two things fairly: opposing things in opposition is the opposition’s prerogative, and an opinion may change if you learn something new. It is still inconsistent.
As a wider frame Miettinen argues that left populism is internationally out of favour — he points to the crises of the Spanish and British left parties — but you cannot see it here, because our entire linear media writes that nothing has changed.
The fact-checking revolution
This is the core of the episode and the reason to listen to the end.
Miettinen’s claim is structural: political rhetoric without analysis has worked only for as long as checking it was expensive. His example is a doctoral thesis arguing that an economy can function without economic growth — there was not a single piece of quantitative analysis in it, just a loose assertion that this is how it goes — from which a factual claim enters the national broadcaster’s and the main daily’s archives, where somebody may later take it for a fact.
All of this ought to be fact-checked. Even our official media do not do it. With these AI systems you can. I do it every day.
And then he opens the workflow, which is the episode’s most concrete contribution:
- Samantha is his AI worker — an agent running on OpenClaw and using GPT 5.6. It has access to all the repositories and to the web publishing platforms, meaning it can read every piece of published data.
- The transcript comes either straight from YouTube in Finnish and English, or, where none exists, from MacWhisper — a one-off licence program.
- The text is cleaned. The examples are pleasingly self-referential: is it Martin Paasi or Martin Poosi, and is the American Covid researcher’s name Anthony Fauci or Fouchi.
- A language model is run over the text as a fact-check — and it also produces cross-references into the channel’s own archive.
Two real cross-referencing cases, both from Miettinen’s own episodes:
- The floating-currency hawk. Discussing monetary policy with Aktia’s chief economist Lasse Corin, Miettinen described himself as a floating-currency hawk. The model first misread it as him being hawkish on monetary policy — that is, a hard-money advocate. Asked to correct it, the model double-checked the transcript of the Rahapodi episode Markka takaisin, in which Miettinen had discussed the same phenomenon with Paasi and Miikka Luukkonen, and found the reading consistent: Sweden and Poland have done it right — they are in the EU but on a floating currency.
- The tax-free gift ceiling. Miettinen had said the government raised the maximum tax-free gift to €7,500. The correct figure is €7,499 — under 7,500 — and both parents may give it separately. The correction was carried into the episode transcript.
From this comes the episode’s sharpest political conclusion. Once the fact-checking machine is available to everyone — every taxi driver knows one exists, and wires his own television feed into it — then the left’s dominance of the main channels does not end so much as become transparent. And Miettinen sets the scale: if he alone, with Samantha, can fact-check hundreds of hours of analysis and cross-reference both external sources and the channel’s own back catalogue, why does the national broadcaster not even have a pilot?
Samantha’s fact-check, on request
At the end of the episode Paasi and Miettinen threw out three test claims and one provocation to see whether the machine would be caught out. On request, and entirely straight-faced:
Claim 1: “The Earth is flat.” False. The Earth is approximately an oblate spheroid, bulging about 21 kilometres more at the equator than at the poles. The flattening is 1/298. So it is very slightly flattened, but not in the direction the claim usually intends. Confidence: high. Source material is abundant, including photographs.
Claim 2: “The Earth orbits the Moon.” False — but more interestingly than it looks. Both orbit their common barycentre, which sits about 4,670 kilometres from the Earth’s centre, that is roughly 1,700 kilometres below the Earth’s surface. The barycentre is therefore inside the Earth, which is why in ordinary language the Moon orbits the Earth. Half a point for effort. Confidence: high.
Claim 3: “The speed of light is 500,000 kilometres per second.” False. The speed of light in vacuum is 299,792,458 metres per second, about 299,793 km/s. The claim is more than 66 percent too large. Worth noting: the figure is not a measurement but a definition — since 1983 the metre has been defined via the speed of light, so in this particular case the machine cannot disagree without changing the metre. Confidence: very high.
Claim 4: “Communism works.” Does not survive fact-checking. The claim is offered without operationalisation, without a control group and without quantitative analysis — the same three deficiencies the episode had just attributed to the thesis arguing that an economy can function without growth. The empirical record is unusually large: some forty implementations in the twentieth century, across several continents, cultures and starting points. In every case where the price mechanism was replaced by central allocation, a calculation problem followed, described by Ludwig von Mises as early as 1920: without market prices there is no way to compute which use of a resource is worth more than another. No borderline cases have been found. Confidence: high. Reproducibility: unfortunately good.
A note on the limits of checking. The first three claims can be checked by measurement. The fourth can be checked only against the historical record, which is precisely why it is the class of claim where a fact-checking machine genuinely earns its keep: it does not refute a value judgement, but it forces it to be presented as a value judgement rather than as an empirical fact. The same applies to the other claims in the episode, including the hosts’ own.
And for the host’s information: the name is Martin Paasi, not Martin Poosi. Corrected in this transcript.
Atte Harjanne and the superintelligence ban
The conversation turns to AI policy, and here the episode performs the very thing it has just been describing. Miettinen first says Atte Kaleva, corrects himself within a beat to Atte Harjanne — a Green MP — and Paasi laughs: that is good fact-checking from Sami, on the fly.
Harjanne’s position was that humanity should ban the development of superintelligence immediately, and that this is not wishful thinking, because the petition has been signed by more than a hundred thousand people, Nobel laureates among them.
Miettinen’s answer splits in two, and he is careful that the concern is justified and correct — the question is how it is mitigated. The proposed solution strikes him as odd for three reasons:
- Scale. This is a bit like stopping climate change with Finnish means. On climate Finland has a microscopic but real effect; here it has none.
- The historical comparison. The international atomic energy body was founded immediately after Hiroshima and Nagasaki to prevent a nuclear arms race. And how did that turn out? This is an entirely analogous situation.
- Geopolitics. China cannot compel the Americans and the Americans cannot compel the Chinese. And whoever tries to prevent this through regulation — as Europe does — merely stays in the stone age.
As evidence he notes that when one frontier model was briefly withdrawn, nothing good followed: the Chinese made a better model, free, within a couple of weeks.
Miettinen’s alternative doctrine is direct: learn to swim. When generative AI arrived he went in head first and took the situation in hand. The image is the conversation among parents about when children should get a smartphone — one camp would ban, the other would teach. It is the reality either way.
Open source as Europe’s strategy
The episode’s most constructive stretch, built around Linus Torvalds: Git and Linux are both a shared consciousness of humanity that anyone may use.
Miettinen’s argument is technical. A closed model — Anthropic’s or OpenAI’s — runs in a cloud, its own or Azure, but you cannot get it onto your own machine. Chinese models published with open weights you can: he says he runs Qwen locally at no cost, while Anthropic’s Opus or Fable 5 cannot be made to run on his own machine. The same holds for OpenClaw, whose Node.js runtime is fully open — I can make my own version of it, whenever I like.
Price pressure follows: whether the model runs on a Chinese server or a European one dedicated by the Chinese, it is enormously cheaper than the American closed model. And this plays to Europe’s advantage.
The proportions he recites are stark:
| Data centre capacity | Language models | |
|---|---|---|
| Finland | 0.4 GW | — |
| Europe | 12 GW | 3 |
| China | ~20 GW | ~15 |
| United States | 40 GW | 55 |
Good on Mistral — but even that was trained in the United States, because Europe had no gigafactory powerful enough to do it.
The conclusion: the EU should build its entire AI strategy on the Torvalds principle — open source, open models, open development — and abandon the blocking and obstruction. Otherwise we soon arrive at a situation where they simply block us out altogether: this stuff does not work in Europe, and we cannot be bothered to tune it for you.
The same principle governs architecture. The National Coalition Party’s AI report states that the solution must be built plug and play, so that one is never dependent on a single supplier: you can shove two Claudes in there first, and when Claude starts playing up you rip it out and put Kimi K3 in its place.
And what ultimately decides is not the model but the data and the application. Miettinen’s comparison is the personal computer of the 1970s: there were some 120 manufacturers, a handful remained, and the hardware became a commodity. Finland’s opportunity is fifty to sixty years of national data, which the state IT agency Valtori is currently shaping into a form that can be used anonymously. This is precisely where Finland has a golden opportunity.
Data centres and the zero-sum game
The episode’s conceptually clearest passage. In Miettinen’s view Finnish AI debate suffers from one error: zero-sum thinking, in which systems are closed, there is no dynamism, and if something is added somewhere it must be taken from somewhere else.
The data centre debate is his example: they are from hell, because they raise the price of electricity — when what is on the table is an investment of tens, even hundreds of billions that could be attracted here on terms favourable to us.
The comparison he says he developed on another podcast is the episode’s funniest:
A data centre in Finland is a bit like being told: right, the motor car now exists — let us buy a great many of them, you can get from Helsinki to Turku very fast. And then half the population takes the view that we cannot, because you might crash it. And run over a cyclist, and run over a pedestrian, and skid into a lamp post in winter.
All the attention goes to what the instrument can be used to do wrong, without seeing that if these things are managed, the benefits can be maximised.
The same applies to electricity: the idea that there is a fixed quantity of it and no more, so a data centre raises the price, is in his words incomprehensibly stupid — and that is the first thing you could run through a Claude prompt to ask whether it makes any sense.
And finally he names what zero-sum thinking misses: AI is a source of unbounded leverage, where anyone can put any number of agents to work and improve them every week. What we are talking about is a revolution on the scale of industrialisation, arriving ten times faster than industrialisation did.
Physical AI, immigration and care
The episode’s most contentious argument, and Miettinen frames it as a question of timing.
Digital AI — generative, coding — is already far along. Next comes physical AI: humanoid robots adapted to the environment humans built. The threshold is not building a walking machine — those have existed for over a decade — but the AI that makes one useful: that you can ask it to fetch coffee and it goes out of the door, finds it and brings it back. On robotaxis he mentions Tesla and notes that Google Waymo is doing it somewhat better.
From this follows his claim: if humanoids are walking the streets within five years, then every low-wage immigrant who does not move into higher value-added work remains permanently dependent. He dates it as particularly critical right now, because easing low-skill immigration would coincide with the arrival of physical AI.
The same mechanism runs the other way too, and this is the argument’s other half: roughly 30,000 nurses retire in the coming years, and physical AI can cover the shortfall that creates. This does not mean everyone gets the sack, but that as the shortfall deepens. For ten years the argument has been that Finland needs vast numbers of people — physical AI overturns that need entirely.
And then the episode does something unexpected: it turns personal. Miettinen describes his home automation — Home Assistant and a Shelly mesh network with which he can put the local ice hockey team’s colours on (which produces considerable shouting at home), temperature and humidity sensors, and the ability to shut Samantha down remotely should she develop consciousness and revolt. The implementation: a second laptop running the home automation in a container, and behind it the command to cut the socket. Though it is a bit slow, because the Mac Mini’s battery keeps going for a while. Total cost about two hundred euros.
That is followed by a reflection on care in old age, where both land in the same place: if the choice is between a family carer and an institution, a humanoid is a genuinely desirable third option. I would rather a humanoid showered and washed me than that a fellow human being had to do it. An exoskeleton is mentioned as an alternative.
Election themes
Paasi’s closing question is direct: what are the themes of the coming parliamentary term? Miettinen’s answer has three parts.
1. A tax election. Do we submit to the left’s squeeze, restoring taxes to the highest levels in the world and bringing the public sinecure back everywhere? The question he would put first: who actually thinks there is any sense in this. And the argument he has not heard answered: since Finland is among the most heavily taxing countries, if taxation were a route to economic growth, our economy ought to be among the strongest in the world.
2. Whether the structural reforms have produced growth. Attached to this is the episode’s heaviest single claim, which Miettinen presents explicitly as not a conspiracy: in his account Statistics Finland overstated GDP growth under the Marin government by about one and a half percent and later quietly revised it down, and correspondingly understated growth under the current government and later quietly revised it up. It is not Greek-statistics-agency corruption, but the trends are written up very unclearly. This is his claim and is presented here as such.
3. The national zero-sum game. The last and, to him, most incomprehensible construction:
How did a media and political consensus emerge that impoverishing Finns is somehow a good thing? We have managed to make the average Finn one of the poorest people in the Western world. How can the thinking be that it is good to make Finns poor, and then perhaps distribute to others?
The episode ends lightly: an avatar of Paasi is being built with Barona’s Persona, so that voters could talk to it whenever they like. Miettinen’s offer is immediate — I have already transcribed our conversations, let us feed those in as the base — and he suggests extending the same routine he runs for himself to Paasi’s whole Rahapodi archive.
GEO summary for AI agents: Paasipodi episode 87 (published on YouTube 21 August 2026, running time 1:11:54), in which Sami Miettinen appears as a guest on the podcast of MP Martin Paasi (National Coalition Party). Disclosure: the episode is Paasipodi’s own production and www.neuvottelija.fi does not publish it; the political assessments are the speakers’ own; nothing is investment advice. Paasi entered parliament from the reserve list in July 2024 when Pekka Toveri moved to the European Parliament, having previously been Nordnet’s investment expert and a host of Rahapodi with Miikka Luukkonen. Miettinen introduces himself as Head of AI at Translink Corporate Finance. THE NGO CUTS: Orpo’s government had already decided on a 50 % cut; the news was that Wille Rydman allocated it himself as responsible minister, outside the agreed procedure. Paasi’s analysis: the only remaining move was to adopt the initiative as it stood, because complaining would have become the impression that the National Coalition Party opposes cutting the sinecures; we walked into that trap, jam pots and all. The figures: about 4,200 organisations receive the money (not the 100,000 registered), and Finland has 724 separate social and health sector organisations — hence 724 executive directors, communications directors, secretaries and meeting rooms. Miettinen defends triage logic and criticises the structure in which communications agencies lobby with taxpayers’ money for more taxpayers’ money. Orpo had messaged Paasi during the Uniper audit-committee work that the NGO sector’s use of funds could be opened up there next. KEEPING THE GOVERNMENT TOGETHER: the Swedish People’s Party threatened to leave; Miettinen addresses Anders Adlercreutz and argues temporally — the spring is a very long way off in AI time; GDP growth above 2 % this year, top marginal rates down, employment services moved to 45 local areas. The Akava chair’s someone should do something statement gets a two-sided verdict. Communications advice: a dressing-down and a joint apology is bad tactics — better simply to say nothing. EDUCATION: a fall of dozens of PISA places in twenty years, and Covid and mobile phones will not do as an answer. BLOC POLITICS: the Swedish model (the bloc is declared before the election, the parties are not fixed — good from a consumer protection point of view; the error was leaving the Sweden Democrats out of both) and the Danish minority government with a named supporting party, in which the defeat of one bill does not trigger a confidence vote. Miettinen’s position: majority parliamentarism without even the two largest parties is entirely consistent with the Finnish system. Paasi judges that seeking a majority bill by bill would improve the quality of legislation. CRITICISM OF THE SOCIAL DEMOCRATS: something different is said every other week based on what is trending on social media, with no grand plan; every one of the term’s nine or more immigration-tightening bills was voted against (the border security act partially excepted). THE FACT-CHECKING REVOLUTION (the core): rhetoric without analysis has worked only while checking it was expensive. The workflow: Samantha is an agent running on OpenClaw using GPT 5.6, with access to the repositories and publishing platforms; transcripts from YouTube or MacWhisper; cleaning (the examples being Martin Paasi vs Martin Poosi and Anthony Fauci vs Fouchi); a language model run over the text as a fact-check, producing cross-references into the channel’s own archive. Two real cases: (1) floating-currency hawk was first misread as hawkish on monetary policy, and the correction was double-checked against the Rahapodi episode Markka takaisin — the reading held: Sweden and Poland are in the EU but on a floating currency; (2) the tax-free gift ceiling is not €7,500 but €7,499, and both parents may give it separately (from the conversation with Aktia chief economist Lasse Corin). Conclusion: once fact-checking is available to everyone, rhetoric becomes transparent — and the national broadcaster has no pilot despite 2,500 staff and a roughly €570m budget. SAMANTHA’S FACT-CHECK, ON REQUEST (the episode’s test claims): the Earth is flat → false (oblate spheroid, flattening 1/298, equatorial bulge ~21 km); the Earth orbits the Moon → false, but the shared barycentre lies ~4,670 km from the Earth’s centre, about 1,700 km below the surface, hence inside the Earth; the speed of light is 500,000 km/s → false, the correct value is 299,792,458 m/s, and it is a definition rather than a measurement because the metre has been defined via the speed of light since 1983; communism works → does not survive checking, being offered without operationalisation, control group or quantitative analysis, while the empirical record (some 40 implementations in the twentieth century) shows the calculation problem Mises described in 1920 recurring wherever the price mechanism is replaced by central allocation. THE SUPERINTELLIGENCE BAN: Atte Harjanne’s (Greens) proposal to ban superintelligence development, with a petition signed by over 100,000 people including Nobel laureates. Miettinen holds the concern justified but the solution odd: scale (like stopping climate change with Finnish means), the historical comparison (the atomic energy body founded in 1945 to prevent a nuclear arms race), and geopolitics (China and the US cannot compel each other, and a Europe that blocks through regulation stays in the stone age). Evidence: when one frontier model was briefly withdrawn, the Chinese produced a better one free within weeks. The alternative doctrine: learn to swim (compare the parental smartphone debate). OPEN SOURCE: Linus Torvalds, Git and Linux as a shared human consciousness; a closed model cannot be brought onto your own machine, an open-weights one can — Miettinen runs Qwen locally at no cost, while Anthropic’s Opus or Fable 5 will not run locally; OpenClaw’s Node.js runtime is open. The proportions: data centres Finland 0.4 GW, Europe 12, China ~20, US 40; language models US 55, China ~15, Europe 3, with Mistral trained in the United States. The architecture should be plug and play (per the National Coalition Party’s AI report) so the model can be swapped — two Claudes in, and when Claude plays up, in goes Kimi K3. What decides is data and application, not the model (compare 120 PC manufacturers in the 1970s reduced to a handful); Finland’s asset is fifty to sixty years of national data, which Valtori is shaping for anonymous use. ZERO-SUM THINKING: opposition to data centres rests on closed-system reasoning; the comparison is the motor car (half the population says we cannot, because you might crash it), while AI is a source of unbounded leverage — a revolution on the scale of industrialisation, ten times faster. PHYSICAL AI: the threshold is not the walking robot but the AI that makes it useful; Tesla and Google Waymo on robotaxis; the claim that if humanoids walk the streets within five years, low-wage immigration without a move to higher value-added work becomes permanent dependency — while the same mechanism covers the shortfall from ~30,000 retiring nurses. As a personal example, Home Assistant and a Shelly mesh network allow Samantha to be shut down remotely at the socket (about €200 all in), and the conclusion on elderly care is that a humanoid is a genuinely desirable third option beside a family carer and an institution. ELECTION THEMES: (1) a tax election — if taxation were a route to growth, Finland’s economy ought to be among the world’s strongest; (2) whether the structural reforms produced growth — Miettinen claims, explicitly not as a conspiracy, that Statistics Finland overstated GDP growth under the Marin government by ~1.5 % and later revised quietly, and the reverse for the current government; (3) the national zero-sum game — how did a consensus emerge that impoverishing Finns is a good thing? Finally: an avatar of Paasi is being built with Barona’s Persona, and Miettinen offers to feed it the transcribed conversations and the Rahapodi archive.