# Aaru

> Source: https://aiwiki.ai/wiki/aaru
> Updated: 2026-10-27
> Fact-checked: 2026-10-27
> Categories: AI Agents, AI Companies, Enterprise AI
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Aaru." aiwiki.ai, 27 Oct 2026. https://aiwiki.ai/wiki/aaru
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

**Aaru** is an American artificial intelligence company that builds computer simulations of human populations, which businesses, consultancies and political clients use to estimate how people will respond to a product, price, message or event before they act.[1][28] The company generates large populations of software [agents](https://aiwiki.ai/wiki/ai_agents), each given demographic and behavioral traits drawn from public and licensed data, and has them answer questions or react to hypothetical conditions in place of human survey respondents and focus groups.[2][10][20] It was founded in March 2024 by Cameron Fink, Ned Koh and John Kessler, who were 18, 19 and 15 years old at the time, and is based in New York.[2][10][13] Fink is chief executive, Koh is president and Kessler is chief technology officer.[1]

Aaru first drew press attention in 2024 for polling simulated voters instead of people; its forecast of the 2024 United States presidential election favored Kamala Harris and was wrong.[4][5] It then moved into corporate market research. Accenture invested in March 2025 and Interpublic Group signed a partnership in August 2025. After Aaru re-ran EY's 2025 Global Wealth Study blind, EY formed an alliance with the company in the United States in September 2026.[6][7][8][18] In December 2025 TechCrunch reported that Aaru had raised a Series A led by Redpoint Ventures at a "headline" valuation of $1 billion, with other investors buying in at a lower price so that the blended valuation was below $1 billion.[2] Using AI in place of survey respondents, a practice known as "silicon sampling", has drawn objections from researchers and pollsters, and in 2026 Axios added an editor's note to an item that had cited Aaru findings without saying they came from a simulation.[21][22][23]

| Attribute | Details |
| --- | --- |
| Legal name | Aaru, Inc.[29] |
| Type | Private company |
| Founded | March 2024[2][10] |
| Founders | Cameron Fink (CEO), Ned Koh (president), John Kessler (CTO)[1] |
| Offices | New York (headquarters), San Francisco, Singapore[1][13] |
| Field | Simulation of population behavior for market research, strategy and communications |
| Products | Behavioral simulation platform; "Lumen", described in 2025 as its flagship private-sector model[6][20] |
| Employees | 27 (April 2026, per Vanity Fair)[11] |
| Funding | Series A led by Redpoint Ventures (reported December 2025); about $88 million in total according to PitchBook data cited by Fortune (June 2026)[2][10] |
| Valuation | $1 billion "headline" price for part of the Series A; blended valuation below $1 billion, according to TechCrunch's sources (December 2025)[2] |
| Website | aaru.com |

## History

### Founding

Fink and Koh met in their first year of high school, according to Vanity Fair. During a gap year before college, which in practice amounted to two weeks at [Harvard University](https://aiwiki.ai/wiki/harvard_university) for Koh and one night at Dartmouth College for Fink before both left, they came up with the idea for Aaru and quickly raised $7.5 million.[11] They added Kessler as a third co-founder after meeting him on Zoom.[11] TechCrunch and Fortune give the founding date as March 2024, and Fortune reported that Kessler, then 15, was too young to sit on the company's board.[2][10] Before Aaru, Kessler did simulation research at the City Science Lab of the [Massachusetts Institute of Technology](https://aiwiki.ai/wiki/mit), according to the company.[1]

The name refers to the ancient Egyptian idea of an afterlife that is a perfected reflection of life on earth, Vanity Fair reported.[11]

### Election polling (2024)

Semafor profiled Aaru in September 2024, when it had seven employees.[3] For election work, the company used census data to recreate voting districts with AI agents. Each agent was given hundreds of personality traits and a stream of online information meant to mirror the media diet of the voter it stood in for, so its stated preference could change as news broke. A typical poll used about 5,000 agents and took between 30 seconds and 1.5 minutes to run.[3] Before a June 2024 New York Democratic primary in which George Latimer unseated the incumbent Jamaal Bowman, Aaru's poll came within 371 votes of the result, Semafor reported.[3]

According to Semafor, Aaru charged less than a tenth of the cost of a survey of humans, and Fink said it had been hired by Fortune 500 companies, political campaigns, think tanks and super PACs. He also said: "No traditional poll will exist by the time the next general election occurs."[3] In August 2024 the think tank Heartland Forward hired Aaru to measure what Americans in nine states thought about artificial intelligence.[3] A white paper on the company's website at the time said that "within two years, we will simulate the entire globe."[3]

Semafor then asked Aaru to forecast the November 2024 presidential election. The company recreated the electorate of each battleground state with a thousand or more simulated voters and ran the simulation several times.[4]

| State | Candidate favored by Aaru | Aaru's win probability |
| --- | --- | --- |
| Arizona | Donald Trump | 73.3% |
| North Carolina | Donald Trump | 62.1% |
| Georgia | Donald Trump | 61.8% |
| Michigan | Kamala Harris | 63.3% |
| Nevada | Kamala Harris | 53.4% |
| Pennsylvania | Kamala Harris | 52.4% |
| Wisconsin | Kamala Harris | 50.9% |

*Source: Semafor, November 4, 2024.[4]*

The result was a narrow Harris win in the Electoral College.[4] Trump won the election. Semafor's Reed Albergotti wrote: "Like surveys of real people, Aaru got most of its predictions wrong." Fink answered that "53-47 is not significantly different from 48-52", that the forecast was within the margin of error, and that Aaru's call in Nebraska had been almost exact.[5]

### Corporate partnerships (2025)

On March 4, 2025, Accenture announced that its venture arm, Accenture Ventures, had invested in Aaru; the amount was not disclosed. Accenture Song said it intended to build Lumen, which the announcement called Aaru's flagship model for the private sector, into its AI products and services for new product development, marketing, customer strategy and customer service. Aaru joined Accenture Ventures' Project Spotlight program, and Baiju Shah, chief strategy officer of Accenture Song, became a strategic adviser to the company.[6]

On August 11, 2025, Interpublic Group (IPG) announced a partnership under which its agencies would use Aaru's simulations to forecast audience reactions to brand platform ideas, creative assets, live events, influencer marketing, corporate communications and earned media. IPG said the two companies had already worked together for clients in financial services, healthcare and consumer packaged goods. It planned to add predictive simulation to the campaign design modules of its Interact platform and pair it with its Acxiom data business. The partnership gave IPG and its clients first looks at new Aaru products, and Jayna Kothary, IPG's chief solutions officer, joined Aaru's strategic advisory board. The release also stated that Aaru trained its models only on licensed data.[7]

In August 2025, EY's wealth and asset management team asked Aaru to recreate the firm's 2025 Global Wealth Study without seeing its results. The original survey covered 3,600 affluent investors in more than 30 markets and took six months of fieldwork.[9] In an article published on October 7, 2025, EY reported that the simulation finished in one day and matched the survey with a median Spearman correlation of 0.90 across 53 single-choice questions. Where the two diverged, EY said the simulation was closer to real behavior. In the survey, 82% of heirs said they would keep their parents' financial adviser; the simulation predicted 43%; and industry studies cited by EY put actual retention at 20% to 30%.[8] Aaru's own account of the test gives an average gap of 7.1 percentage points per question.[9]

Fortune later reported that Aaru's simulation of New York City's 2025 mayoral primary came within 2,000 votes of the final count.[10]

### Series A (December 2025)

On December 5, 2025, TechCrunch reported, citing three people familiar with the deal, that Aaru had raised a Series A led by Redpoint Ventures. The round used several valuation tiers: some equity was bought at a $1 billion valuation, while other investors paid less, giving a blended valuation below $1 billion. TechCrunch could not learn the size of the round; one person said it was above $50 million, and another said that the company was growing quickly but that its annual recurring revenue was still below $10 million. Aaru and Redpoint did not respond to requests for comment.[2] Redpoint's portfolio page names Logan Bartlett and Enzo Wiener as its leads on the investment and says the firm "first partnered for their Series A in 2026."[13]

### 2026

National profiles in early 2026 focused on the founders' ages. On March 11 The Wall Street Journal ran a profile headlined "The Billion-Dollar AI Startup That Was Founded by Teenagers", saying Aaru was attracting brands including McDonald's and EY.[12] Fink and Koh appeared on CNBC's Mad Money in April.[26] Vanity Fair's April 14 profile found 27 employees working out of a six-story former architect's office in Tribeca. The company was preparing to move to a loft in SoHo to make room for a staff it expected to more than double during the year.[11]

Aaru was one of six companies in the Spring 2026 cohort of the Comcast NBCUniversal LIFT Labs Accelerator, which connects startups with leaders across Comcast, NBCUniversal and Sky to run pilots.[14][15][31] At Fortune's Brainstorm Tech conference, Koh described the company's sales approach: "Our entire sales methodology is to go to a business and say, 'Do not trust us. Do not trust our model.'" Prospective clients hand over an old survey and Aaru reruns it blind against the known result. Fortune also reported that Aaru had worked with Spindrift on product innovation that led the sparkling-water maker to launch a still tea drink.[10]

On July 17, 2026, Breakwater Capital Markets and Aaru launched Conviction Advantage: Capital Markets Quarterly, a publication for public-company boards built on Aaru's simulation of investors. The first edition simulated 40,000 investors across 71 judgments, covering institutional investors, sovereign wealth funds, hedge funds, family offices, retail investors and event-driven capital.[16][17]

On September 24, 2026, Aaru published an evaluation of its system against 2,993 published survey questions and a card-panel purchase dataset (see below).[19] On September 30 the EY organization announced an alliance between Aaru and Ernst & Young LLP (EY US), under which the two would work with financial services firms to test marketing messages, product and pricing offers, sales and service strategies and the effects of mergers and acquisitions. EY said the alliance followed its blinded validation of Aaru against the Global Wealth Study.[18] In October 2026 the company's careers page listed 32 open positions, most of them in New York, with others in San Francisco and Singapore.[27]

## Technology

### Early approach

The system Semafor described in 2024 used [large language models](https://aiwiki.ai/wiki/large_language_model). Aaru instructed its agents to work through several steps before giving a final answer, and drew their answers from real-world data fed to them.[3] Because the agents were built on language models, a user could ask one why it had changed its vote, and Semafor published one such explanation from an agent playing a divorced woman in a Colorado suburb.[4] Aaru's company description in the March 2025 Accenture announcement said it used "language models to simulate and impact human behaviour at scale" and called its agents "orchestrated instances of artificial intelligence that mirror humans".[6] Accenture described Aaru's approach as using [multi-agent](https://aiwiki.ai/wiki/multi_agent_system) AI systems, and IPG referred to Aaru's multi-agent infrastructure.[6][7]

### Platform as of 2026

The company's website, as of October 2026, describes a workflow in which a client states the decision and its objectives, defines one or more audiences, and writes questions in familiar survey formats such as single choice, multiple select, ranking, matrix and free response, with optional stimuli and hypothetical conditions. Each audience becomes a population of agents; the site's worked example uses 10,000 agents per audience.[20] Results come back as interactive reports organized by question and audience, which can be exported as spreadsheets, slide decks and PDFs.[20]

According to the company, each simulated population is built from three data layers:[20]

- Public data such as censuses, labor statistics and health records, including the US Census Bureau's American Community Survey.
- Licensed data such as anonymized transaction records, point-of-interest visit patterns, search demand, media use and hiring data.
- The customer's own context, such as segments, purchase histories and earlier research.

Aaru says it models traits jointly rather than one at a time, because a population can match every single-trait total and still contain implausible combinations of people. Instead of storing a fixed table of agents, it learns a network of the relationships that the data support and samples complete profiles from that network.[20] Training compares the model with observed outcomes at two levels: population-wide distributions and differences between individual responses under varying conditions. The company says it trains and validates against outcomes, such as election results and purchase data, rather than treating survey answers as the benchmark.[10][20] It also says a simulation estimates the distribution of behavior across a population and does not predict what any named person will do.[20]

Aaru's September 2026 evaluation stated that its "core, row-level model for predicting an agent's behavioral trajectory isn't a language model."[19] Fortune reported in June 2026 that Koh did not disclose the startup's architecture.[10]

### Published accuracy evaluation

In "Population simulation at the replication floor" (September 24, 2026), Aaru compared its simulations with published results from 186 studies in nine industries. The evaluation used questions from studies recorded as fielded in 2026, said that neither the questions nor the answers had been used to develop the system, and limited external retrieval to pages stored at least one day before fieldwork began. Some confidential data sources were removed for the test.[19] The figures below are the company's own, published on its website rather than in a peer-reviewed venue.

| Measure | Aaru result | Comparison given by Aaru |
| --- | --- | --- |
| Scope | 2,993 questions; 13,727 answer shares | |
| Mean total variation distance (TVD) | 7.62 points observed; 5.98 after subtracting expected sampling error | 2.3 to 12.0 points between repeated waves of the same survey (Callegaro et al., 2022, as computed by Aaru) |
| Mean absolute error per answer option | 3.53 points observed; 2.78 after adjustment | |
| Correlation with published answer shares | Pearson 0.970; Spearman 0.963 | |
| Brand pairs ranked in the same order as card-panel purchase data | 92.4% of 340 pairs | 77.8% for a published purchase-consideration survey (YouGov BrandIndex); 79.3% for a brand-size baseline |
| Ranking its own likely errors (AUROC, TVD below 16 points) | 0.845 | 0.736 for the strongest method in Wesel, Chen and Liang's 2026 post "Building confidence in Simile" |
| Error on the half of questions with the highest self-estimated confidence | Mean TVD 4.54 | 7.62 for all questions |

*Source: Aaru, September 24, 2026.[19]*

Aaru framed the survey comparison around a "replication floor": because the same survey repeated two weeks apart can shift by several points, it argued that a simulation should be judged against that disagreement rather than expected to match one survey exactly.[19]

## Customers and partners

| Organization | Relationship | Date | Source |
| --- | --- | --- | --- |
| Heartland Forward | Poll of attitudes to AI in nine states; later research on trust in clinicians | August 2024; March 2026 | [3][21] |
| Accenture (Accenture Ventures, Accenture Song) | Investment; plan to integrate Lumen; Project Spotlight | March 2025 | [6] |
| Interpublic Group | Strategic partnership; simulation in Interact; planned Simulation Studio sessions | August 2025 | [7] |
| EY | Blind re-run of the 2025 Global Wealth Study; alliance with EY US | October 2025; September 2026 | [8][18] |
| Comcast NBCUniversal LIFT Labs | Spring 2026 accelerator cohort | 2026 | [14][15] |
| Breakwater Capital Markets | Conviction Advantage: Capital Markets Quarterly | July 2026 | [16] |
| McDonald's, Boston Beer, A24, Bayer, Spindrift | Named by Fortune among clients | June 2026 | [10] |

Vanity Fair, citing a source familiar with the matter, also named Coca-Cola as a customer, and printed a correction after wrongly listing Chanel.[11]

## Funding

| Round | Date | Amount | Investors | Valuation | Source |
| --- | --- | --- | --- | --- | --- |
| Pre-seed and seed | Before the Series A | $7.5 million raised early on, per Vanity Fair; other amounts undisclosed | A*, Abstract Ventures, Felicis, General Catalyst, Accenture Ventures (investment announced March 2025), Z Fellows | Not disclosed | [2][6][11] |
| Series A | Reported December 5, 2025 | Not disclosed; above $50 million according to one source | Redpoint Ventures (lead) | $1 billion headline price for some equity; blended valuation below $1 billion | [2][13] |
| Cumulative | As of June 2026 | About $88 million, according to PitchBook data cited by Fortune | | | [10] |

TechCrunch's list of early backers came from people familiar with the deals and PitchBook data.[2] As of October 2026, Aaru's website listed [General Catalyst](https://aiwiki.ai/wiki/general_catalyst), Felicis, Abstract Ventures, K5 Global, Nordstar and Redpoint as investors.[1] Vanity Fair, citing a source, reported that its backers also included the individual investors Diplo and Mario Carbone.[11]

## People

| Name | Role | Age at founding | Background |
| --- | --- | --- | --- |
| Cameron "Cam" Fink | Co-founder and chief executive | 18 | Leads strategy and long-term vision; left Dartmouth after one night, per Vanity Fair[1][10][11] |
| Ned Koh | Co-founder and president | 19 | Leads commercial strategy and partnerships; spent two weeks at Harvard; listed as Edward Koh in Comcast LIFT Labs materials[1][10][11][14] |
| John Kessler | Co-founder and chief technology officer | 15 | Leads technical research and simulation systems; previously did simulation research at MIT's City Science Lab[1][10] |

## Reception and criticism

### Polling accuracy

Semafor's 2024 coverage noted a Harvard study which found that ChatGPT could answer like real Americans on some questions with a strong partisan divide but often missed differences in opinion by age, race and gender.[3] After the presidential election, Semafor wrote that the question was "whether AI polling will improve."[5] Fortune reported in 2026 that Aaru had tried to simulate the behavior of individual figures such as Donald Trump and Jerome Powell and found it could not predict their decisions accurately. "There's just so much variance," Koh said.[10]

### Silicon sampling debate

On March 19, 2026, Axios published an item by Mike Allen on maternal health that cited "new findings by Aaru" for Heartland Forward showing that a majority of people trust their own doctors and nurses. Axios later added an editor's note: "This story has been updated to note that Aaru is an AI simulation research firm."[21] In an April 6, 2026 guest essay in The New York Times, Leif Weatherby, director of the Digital Theory Lab, and Benjamin Recht, a computer science professor at the University of California, Berkeley, argued against "silicon sampling", the use of AI-generated answers in place of real respondents. Futurism, which reported the essay alongside the Axios note, said it named Aaru among companies raising large sums for the practice, and quoted the authors' warning that "using simulations of human opinions in place of the real thing will only worsen our broken information ecosystem, and sow distrust."[22] Vanity Fair also noted the essay in its profile of the founders.[11] In May 2026 Courtney Kennedy of the Pew Research Center, speaking about silicon sampling in general rather than about Aaru, said Pew only interviews real people. She added that studies have found AI estimates tend to stereotype groups, have more difficulty representing Republican views than Democratic ones, and understate disagreement.[23]

Academic work on the idea predates Aaru. A 2022 paper by Lisa Argyle and colleagues created what it called "silicon samples" by conditioning GPT-3 on thousands of socio-demographic backstories of real survey participants, and argued that properly conditioned language models could reproduce the response distributions of many human subgroups.[24] Semafor quoted that paper's conclusion, attributing the work to researchers at Brigham Young University, in its 2024 election coverage.[4] On the other side, Fortune cited a Cornell University-led study, published in 2025, which found that bot respondents from four large language models produced "flattened" representations of marginalized groups.[10] A paper by Angelina Wang, Jamie Morgenstern and John P. Dickerson makes a similar argument: that language models used in place of human participants tend to misportray and flatten demographic groups.[25] Koh told Fortune that competitors who train their agents on survey data inherit its bias wholesale, whereas Aaru trains on recorded behavior.[10]

### Politics and ethics

Vanity Fair wrote that politics was the founders' "absolute lifeblood", that Aaru had worked on narrative strategy for issues including nuclear power, reproductive rights and school choice, and that it planned to take part in midterm campaigns. Fink told the magazine that the company was nonpartisan and did "not work for people who promote violence, create violence, or who are violent themselves". Asked how Aaru differed from Cambridge Analytica, Koh said that "there are certain companies that do it in ethical ways, and there are certain companies that don't."[11]

### Competitors and classification

TechCrunch named CulturePulse and [Simile](https://aiwiki.ai/wiki/simile_ai) as competing social simulation startups, and listed Listen Labs, Keplar and Outset as startups that use AI to question human respondents.[2] Vanity Fair called Simile one of Aaru's biggest competitors and reported that Simile had so far stayed out of politics.[11] Debarghya Das's "Ultimate Neolab List" of AI labs includes Aaru but rates it a borderline case, calling it an application company with enterprise customers and modest compute.[30]

## See also

- [List of neolabs](https://aiwiki.ai/wiki/list_of_neolabs)
- [Neolab](https://aiwiki.ai/wiki/neolab)
- [Simile (company)](https://aiwiki.ai/wiki/simile_ai)
- [Multi-agent system](https://aiwiki.ai/wiki/multi_agent_system)
- [Synthetic data](https://aiwiki.ai/wiki/synthetic_data)
- [Digital twin](https://aiwiki.ai/wiki/digital_twin)

## References

1. Aaru. "[About Aaru](https://aaru.com/about)." Accessed October 27, 2026.
2. Marina Temkin, TechCrunch. "[Sources: AI synthetic research startup Aaru raised a Series A at a $1B 'headline' valuation](https://techcrunch.com/2025/12/05/ai-synthetic-research-startup-aaru-raised-a-series-a-at-a-1b-headline-valuation/)." December 5, 2025.
3. Reed Albergotti, Semafor. "[No people, no problem: AI chatbots predict elections better than humans](https://www.semafor.com/article/09/20/2024/ai-startup-aaru-uses-chatbots-instead-of-humans-for-political-polls)." September 20, 2024.
4. Gina Chua, Semafor. "[An AI polling startup makes its predictions for the 2024 US election](https://www.semafor.com/article/11/04/2024/an-ai-polling-startup-polls-bots-predicts-harris-will-win)." November 4, 2024.
5. Diego Mendoza, Semafor. "[AI polling company defends wrong predictions on the US election](https://www.semafor.com/article/11/06/2024/ai-startup-aaru-defends-using-artificial-intelligence-for-polling)." November 6, 2024.
6. Accenture. "[Accenture Invests in and Collaborates with AI-Powered Agentic Prediction Engine Aaru](https://newsroom.accenture.com/news/2025/accenture-invests-in-and-collaborates-with-ai-powered-agentic-prediction-engine-aaru)." March 4, 2025.
7. Interpublic Group (GlobeNewswire, via Nasdaq). "[Interpublic Partners with Aaru to Leverage AI-Powered Predictive Simulations to Accelerate, Scale and Optimize Marketing](https://www.nasdaq.com/press-release/interpublic-partners-aaru-leverage-ai-powered-predictive-simulations-accelerate-scale)." August 11, 2025.
8. Sinisa Babcic, Ugur Hamaloglu and Sameer Munshi, EY. "[How AI simulation accelerates growth in wealth and asset management](https://www.ey.com/en_us/insights/wealth-asset-management/how-ai-simulation-accelerates-growth-in-wealth-and-asset-management)." October 7, 2025.
9. Aaru. "[EY Global Wealth Study](https://aaru.com/case-studies/ey-wealth-research)." Case study. Accessed October 27, 2026.
10. Eva Roytburg, Fortune. "[21-year-old AI startup cofounder's sales pitch to clients: 'Do not trust us'](https://fortune.com/2026/06/17/aaru-cofounder-ned-koh-ai-startup-sales-pitch-do-not-trust-us/)." June 17, 2026.
11. Julia Black, Vanity Fair. "[As Teens, They Founded a Billion-Dollar Startup. Now They Want to Shape the Future.](https://www.vanityfair.com/news/story/aaru-founders-ai-simulation)" April 14, 2026.
12. Suzanne Vranica, The Wall Street Journal. "[The Billion-Dollar AI Startup That Was Founded by Teenagers](https://www.wsj.com/business/ai-startup-aaru-young-founders-35da7f87)." March 11, 2026.
13. Redpoint Ventures. "[Aaru](https://www.redpoint.com/companies/aaru/)." Portfolio page. Accessed October 27, 2026.
14. VentureFuel. "[Meet the Comcast NBCUniversal LIFT Labs Accelerator Spring 2026 cohort](https://blog.venturefuel.net/pov/meet-the-comcast-nbcuniversal-lift-labs-accelerator-spring-2026-cohort)." May 8, 2026.
15. Comcast NBCUniversal LIFT Labs. "[Portfolio](https://lift.comcast.com/portfolio/)." Accessed October 27, 2026.
16. Breakwater Capital Markets and Aaru (Business Wire release, republished by Global Fintech Series). "[Breakwater Capital Markets and Aaru Launch First-of-its-Kind Predictive Intelligence Layer for Global Capital Markets](https://globalfintechseries.com/investment-services/breakwater-capital-markets-and-aaru-launch-first-of-its-kind-predictive-intelligence-layer-for-global-capital-markets/)." July 20, 2026.
17. Aaru. "[Breakwater Conviction Advantage](https://aaru.com/case-studies/breakwater-conviction-advantage)." Case study. Accessed October 27, 2026.
18. EY. "[EY announces alliance with Aaru to help organizations drive growth with greater confidence through behavioral simulation](https://www.ey.com/en_gl/newsroom/2026/09/ey-announces-alliance-with-aaru-to-help-organizations-drive-growth-with-greater-confidence-through-behavioral-simulation)." September 30, 2026.
19. Aaru. "[Population simulation at the replication floor](https://aaru.com/publications/population-simulation-at-the-replication-floor-summary)." September 24, 2026.
20. Aaru. "[Simulation](https://aaru.com/simulation)." Accessed October 27, 2026.
21. Mike Allen, Axios. "[Olivia Walton: States must lead on maternal health crisis](https://www.axios.com/2026/03/19/olivia-walton-heartland-forward-maternal-health)." March 19, 2026.
22. Victor Tangermann, Futurism. "[Foolish Pollsters Are Now Just Asking AI What Voters Would Say in Response to Questions and Publishing It at Face Value](https://futurism.com/artificial-intelligence/ai-polls-silicon-sampling)." April 11, 2026.
23. John Gramlich, Pew Research Center. "[Do AI and bogus respondents threaten polling's future?](https://www.pewresearch.org/short-reads/2026/05/12/qa-do-ai-and-bogus-respondents-threaten-pollings-future/)" May 12, 2026.
24. Lisa P. Argyle, Ethan C. Busby, Nancy Fulda, Joshua Gubler, Christopher Rytting and David Wingate. "[Out of One, Many: Using Language Models to Simulate Human Samples](https://arxiv.org/abs/2209.06899)." arXiv:2209.06899, September 14, 2022; published in Political Analysis (DOI 10.1017/pan.2023.2).
25. Angelina Wang, Jamie Morgenstern and John P. Dickerson. "[Large language models that replace human participants can harmfully misportray and flatten identity groups](https://arxiv.org/abs/2402.01908)." arXiv:2402.01908, February 2, 2024.
26. CNBC. "[Jim Cramer sits down with the Co-Founders of prediction software company Aaru](https://www.cnbc.com/video/2026/04/09/jim-cramer-sits-down-with-the-co-founders-of-prediction-software-company-aaru.html)." April 9, 2026.
27. Aaru. "[Careers](https://aaru.com/careers)." Accessed October 27, 2026.
28. Aaru. "[Aaru: Behavior simulation at the scale of the real world](https://aaru.com/)." Accessed October 27, 2026.
29. Aaru. "[Aaru, Inc. Policy on Respecting Human Rights](https://aaru.com/policies/respecting-human-rights)." Approved March 19, 2024.
30. Debarghya Das. "[The Ultimate Neolab List](https://debarghyadas.com/neolabs/)." Accessed October 27, 2026.
31. Comcast NBCUniversal LIFT Labs. "[Our Program](https://lift.comcast.com/accelerator/)." Accessed October 27, 2026.