Hiasynth Builds a Synthetic Europe Small Firms Can Query

Hiasynth has raised an undisclosed angel round to launch a 1:1 synthetic model of Europe’s population, half a billion personas across 37 countries. Further Than Capital led, three more angels joined, and founder Emma Agering plans to show the product in September 2026.

The pitch is blunt. If the continent is one model instead of 37 research projects, a ten-person company can read a market across Europe as cheaply as the street next door. The catch the announcement does not dwell on is that synthetic people are already a product, and Germany already has a near-complete virtual census you can talk to.

How Hiasynth Built a Queryable Europe

Hiasynth says each persona carries more than 600 attributes that move together as they do in real life, including income, housing, consumption, habits and values. None of them is a real person. Agering’s method is to refuse isolated profiles. She mapped 3.7 million micro-zones, built the households, and placed every persona in one, on the argument that where you live shapes lifestyle as much as who you are.

THE MODEL IN FOUR FIGURES

• Headcount: Half a billion synthetic personas, described as a 1:1 model of Europe.
• Map: 37 countries, each with its own income levels, tax systems and cultural values.
• Traits: More than 600 attributes per persona that are built to co-vary, not sit as a static list.
• Place: 3.7 million micro-zones, or about 135 people per zone if the headcount is taken as 500 million.

The public site frames that stack as a synthetic population of Europe for AI and greets visitors with named sketches such as Finn, Lottie, Raymond, Macy, Braam, Tung, Maggie, Fred and Klaus. It also says the model is not a still photo. Markets are projected forward with fertility, income paths, migration and household formation, which matters on a continent where Eurostat counted population falling in 11 EU countries during 2025 and rising in 16.

Households, Not Isolated Profiles

“Where and how you live define your lifestyle as much as who you are as an individual. So we mapped 3.7 million micro-zones, built the households, and placed every persona inside one. That is what makes the consumption patterns real,” Agering said.

That household step is the product’s claimed edge over a chatbot asked to “act like a Spanish millennial.” A language model can invent a shopper. It cannot, on its own, keep that shopper consistent with the building she lives in, the tax table she faces, and the people who share her kitchen.

Values, Money, and a Forward Clock

The site lists three layers behind the answers: values (trust, priorities, outlook), segments that cross borders rather than stopping at national dashboards, and a clock that runs past the present. Design partners are still being invited at hello@hiasynth.co, which is a useful tell. The angel round funds launch and product work. The model is not yet a self-serve panel with a public price list.

Europe Still Pays for Research Country by Country

Agering’s commercial claim is that market intelligence has been a luxury priced for enterprise, and that the price is about to collapse. She points to a habit European companies know too well. Each country brings its own language, its own consumption patterns and its own research invoice, which is why so many firms have opened in the United States before they opened next door.

The invoice is not folklore. Industry figures circulating among insights vendors still put a typical custom study at four to eight weeks and $15,000 to $65,000. In the DACH region, classic pre-testing is often quoted at €8,000 to €60,000 a wave, against €600 to €8,000 for a synthetic run of the same brief. Bain has described synthetic-customer work, when it sits on real research data, as taking half the time and about one-third the cost of the old method.

WHAT A STUDY STILL COSTS

The old panel machine is also tiring. Response rates on online panels are described as sliding from 20 to 25 percent in 2019 to 10 to 15 percent by 2025, while fraud and bot noise keep climbing. A 2025 Qualtrics study found 62 percent of market researchers had used synthetic data in the prior six months, and 87 percent of those users called the results positive. Gartner has forecast that 60 percent of product marketing teams will use synthetic customer personas by 2028, up from 5 percent in 2025.

None of that proves Hiasynth’s price will fall through the floor for a ten-person shop. It does show why an angel cheque for a queryable population is not a science-fair bet. The research bill is the thing European expansion has always tripped on, and the tools that claim to flatten it are already in market.

Germany Already Runs 83.6 Million Virtual Residents

Call the Hiasynth round a category invention and you miss the shelf it is walking onto. Statista already sells 83.6 million virtual German residents, each with 40-plus consumer attributes, located on a 100 metre by 100 metre grid. The firm stresses that no generative model writes the people, that official statistics and its own surveys are fused by human modelling, and that the output is meant to be GDPR-safe because no real person is on the screen. A sample prompt on that product is ordinary and sharp: size the market for electric vehicles among affluent professionals in their early 30s, or test an outerwear shop around Savignyplatz in Berlin.

That German grid is the awkward neighbour. Eurostat put the EU population of 452.0 million on 1 January 2026, with Germany at 83.5 million. Hiasynth’s half a billion across 37 countries is wider than the Union, which only works if the United Kingdom, Norway, Switzerland and other non-EU markets sit inside the same file. The 1:1 claim is a scale claim, not a claim that nobody else built a synthetic person.

WHO ALREADY SELLS SYNTHETIC PEOPLE

GWI, the consumer research firm, draws a line the rest of the category now recites. A synthetic persona is useful when it is grounded in consented survey data you can cite. A chatbot asked to play a demographic is a stereotype with nice manners. GWI says its own layer sits on more than 2 million surveys a year across 50-plus markets. Calibrated tools in vendor tests claim 85 to 95 percent parity with human panels on concept, price and message work. Generic prompts without that calibration sit closer to 55 percent.

Hiasynth has not published a head-to-head of that kind. Until it does, the 600 attributes and the household graph are a method, not a score. The honest reading of the shelf is that Agering is not inventing synthetic Europe. She is trying to make the whole continent as queryable as Statista has already made Germany, and to do it at household grain rather than a 40-attribute slice.

The Population Will Answer Through Claude and ChatGPT

Distribution is the other half of the bet. Hiasynth will not ask a ten-person company to learn a new insights platform. It will ship through MCP, the open protocol for model tools that lets assistants such as Claude and ChatGPT call outside data and tools over JSON-RPC. GWI already uses that door for its Agent Spark personas. Other synthetic-research servers have listed MCP connectors for the same reason. The interface war is largely over. The fight is over whose population the assistant is allowed to trust.

WHAT MCP CHANGES IN PRACTICE

• No extra dashboard: A strategist can ask the model in the chat tool they already use, rather than booking a portal seat.
• Tool, not essay: The assistant is supposed to query a structured population instead of improvising a market from web text.
• Same pipe, many vendors: MCP is an open standard, so Hiasynth will sit next to other data servers, not above them.
• Day-one Europe: Björklund’s phrase is that the product ships across the continent on day one, which only holds if the 37-country file is live behind that pipe.

Stefan Lindeberg, who sits at Further Than Capital and co-founded Creandum, put the failure mode in one run of sentences. Language models are fluent. They are not a census.

Language models are good at producing a plausible answer about any market and bad at knowing whether that answer is true. Hiasynth replaces the guess with a population the model can actually reason over. That is the difference between an AI that sounds convincing and one you can act on.

Stefan Lindeberg, Further Than Capital

The irony sits in that paragraph. The proposed cure for an AI that invents a market is a continent of invented people. It only works if the invented people are constrained by the same joints as the living ones, including housing, tax and the dull geography of a micro-zone. If they are not, Hiasynth becomes another fluent guess with better branding.

Further Than Capital Wants in Before the Seed

The cheque is an angel round, not a seed, and the amount was not published. Further Than Capital led. Three unnamed angels joined. The money is earmarked for launch and product work ahead of a planned seed. Per Björklund, at Further Than Capital, sold the decision as a founder bet rather than a category bet.

“We back founders before ideas, and Emma has built startups before. When the product ships across Europe on day one, we want to be in from the start,” Björklund said. Public records do not spell out those earlier companies, so that line should be read as the investor’s diligence claim, not as a filed biography.

Further Than Capital’s own site is a long portfolio wall. The announcement highlights early positions in Lovable, the software builder, and Blackwall, a site-protection firm. Lindeberg’s profile places him in Stockholm County. Nordic tech lists treated the Hiasynth note as a Stockholm-adjacent item, which fits the names around the table even though Hiasynth has not posted a city on the homepage.

WHAT WE KNOW

• The round: Angel financing led by Further Than Capital, with three additional angels, amount undisclosed.
• The file: Half a billion personas, 37 countries, more than 600 attributes, 3.7 million micro-zones.
• The door: Launch in September 2026 through MCP, first public outing at Bits and Pretzels in Munich.
• The next cheque: A seed round is planned; the live site still invites design partners and early investors.

WHAT IS UNCONFIRMED

• Price: No public tariff for a ten-person company, despite the collapse-in-price claim.
• Proof: No released accuracy test against live European surveys or panels.
• The map: No published list of the 37 countries, and no method paper for how attributes co-vary.
• The founder file: Prior startups named by the lead investor, not by a public company register in the announcement.

Privacy is the quiet reason a product like this exists. A true 1:1 file of living Europeans, with 600 linked traits and a household key, would be a GDPR nightmare. Virtual people are the legal workaround. They are only a workaround if the generator cannot be reverse-engineered back to anyone real, and if the training trail itself had a lawful basis. Hiasynth has not put that paperwork on the homepage either.

Munich Sees the Product Before Paying Customers Do

Hiasynth will launch in September 2026 at Bits and Pretzels, the Munich founders festival in late September that runs from 28 to 30 September 2026 at the ICM on the Messe München grounds. Organisers pitch about 7,500 founders, investors and operators, with the third day spilling into Oktoberfest. It is a loud room in which to say that Europe is now one query. It is also a room full of people who have already been sold an MCP plugin this year.

FROM ANNOUNCEMENT TO A HALL IN MUNICH

1. 2 September 2026: Hiasynth announces the angel round and the 37-country model.
2. September 2026: Product work continues on the back of that capital, with a seed still ahead.
3. 28 to 30 September 2026: Planned launch at Bits and Pretzels in Munich, with MCP access for assistants such as Claude and ChatGPT.

Between those dates the company still has to show that a persona in Lisbon and a persona in Lille do not collapse into the same fluent European. It has to show that 3.7 million micro-zones are more than a marketing integer. And it has to pick a price that matches Agering’s line about a ten-person company, because enterprise insights software does not collapse by being announced.

The round buys a launch, not a verdict. Agering will walk into Munich with a continent of people who do not exist, and ask operators to trust them more than a confident chatbot. If the households hold, a small firm might finally read Italy the way it reads its own postcode. If they do not, Europe will have gained another plausible answer about itself, at 1:1 scale.