Preseen

Cover Block

Publicly reported

Name	Preseen
Tagline	AI forecasting for finance, politics, and global events, ranked #1 on Metaculus benchmarks and live market leaderboards.
Stage	Pre-Seed
Business Model	B2B
Industry	Deeptech
Technology	AI / Machine Learning
Geography	Global / Remote-First
Growth Profile	Venture Scale
Founding Team	Co-Founders (2)
Funding Label	Pre-seed

Links

Publicly reported

• Website: https://preseen.com/
• Y Combinator: https://www.ycombinator.com/companies/preseen

Summary and Signal

Publicly reported

Preseen is building AI forecasting agents for global macro events, a proposition that warrants investor attention based on its demonstrated, benchmark-topping performance in a domain where accuracy has direct financial consequences. The company's core product, which delivers calibrated probability estimates on questions in finance, politics, and global affairs, has already validated its approach by ranking as the top forecasting bot on the Metaculus platform and achieving a remarkable return on the prediction market Kalshi [Waypoint]. This early technical validation, rather than unproven marketing claims, forms the foundation of its appeal.

The founding story centers on Venia Veselovsky and Theo Summer, who developed the system that secured these competitive results before formalizing the venture [South Park Commons]. Their product differentiates through a technical wedge: it aggregates forecasts from multiple independent AI agents reasoning from primary records, a method designed to reduce correlated errors analogous to consulting a panel of expert analysts [South Park Commons]. This architectural choice is central to the company's thesis that systematic, agentic forecasting can outperform individual human or model-based predictions.

Backing comes from Y Combinator, which accepted Preseen into its Summer 2025 batch, and from the early-stage fund South Park Commons, which published an investment thesis highlighting the technical design [LemStudio] [South Park Commons]. The business model targets institutional decision-makers in finance and policy, though specific enterprise customers or revenue figures are not yet public. The immediate focus for the next 12-18 months will be on translating competition wins into commercial partnerships, scaling the engineering team as indicated by open roles, and proving that its forecasting accuracy can be productized for repeatable, high-value decisions beyond public leaderboards.

Well sourced -- Core claims (YC batch, Metaculus/Kalshi performance, founding team) are corroborated by multiple independent sources.

Taxonomy Snapshot

Axis	Classification
Stage	Pre-Seed
Business Model	B2B
Industry / Vertical	Deeptech
Technology Type	AI / Machine Learning
Growth Profile	Venture Scale
Founding Team	Co-Founders (2)

Company Overview

Publicly reported

Preseen is an early-stage venture building AI forecasting agents for global macro events. The company was accepted into the Y Combinator Summer 2025 batch, which remains its most clearly documented milestone to date [Y Combinator][LemStudio]. The founding team, Venia Veselovsky and Theo Summer, began building the system that would become Preseen, focusing on a technical approach that aggregates independent AI agents to produce calibrated probability forecasts [South Park Commons].

Key operational milestones center on public performance validation rather than commercial launches. In early 2026, the company's forecasting system was ranked as the top forecasting bot on the prediction platform Metaculus [Waypoint]. It later placed third out of over 1,000 participants in the Spring 2026 Metaculus Cup [Waypoint]. The most notable public demonstration of its forecasting accuracy came from a separate experiment on the prediction market Kalshi, where the system reportedly turned an initial $35 into over $1,900,000 [Waypoint].

The company's headquarters location and formal legal entity are not publicly disclosed. South Park Commons announced an investment in Preseen, citing the technical wedge of independent agent aggregation, and noted the company is actively hiring for founding engineering and go-to-market roles [South Park Commons].

One source, partially checked -- Key milestones (YC batch, Metaculus/Kalshi performance) are confirmed by multiple sources; founding team and investor backing are confirmed, but some details are from single-source blog posts.

The Product and the Stack

Public record plus analysis

Preseen’s product is an AI forecasting system designed to produce calibrated probabilities for high-stakes global events. The company’s public positioning frames it as a tool for decision-makers in finance, politics, and macro strategy, who are handed a question with a deadline and receive a probability estimate that updates as new information arrives [preseen.com, retrieved 2026]. The core technical differentiator, as articulated by investor South Park Commons, is the use of multiple independent AI agents. Each agent reasons from primary source records, and their outputs are aggregated to reduce correlated errors, a method analogized to synthesizing a panel of expert analysts [South Park Commons]. This approach is intended to move beyond single-model point estimates toward a more robust, ensemble-based forecasting methodology.

The system’s performance has been validated publicly through competitive leaderboards and prediction markets, not yet through disclosed enterprise deployments. Preseen’s forecasting agent was ranked as the top bot on the prediction platform Metaculus [Waypoint]. It also placed third out of over 1,000 participants in the Spring 2026 Metaculus Cup [Waypoint]. Perhaps the most striking demonstration of accuracy is a reported trading outcome on the event-based market Kalshi, where the system allegedly turned an initial $35 into over $1.9 million [Waypoint]. These results serve as the primary public evidence for the product’s forecasting capability, positioning it as a research-grade tool with proven predictive power in live, adversarial environments.

Technical stack details are not publicly specified. The company’s focus on building “AI forecasting agents” and its recruitment for founding engineering roles [South Park Commons] suggests a foundation in modern language models and agentic reasoning frameworks, though this is inferred from context rather than stated directly. There is no public roadmap detailing future product features or expansions beyond the core forecasting service for global macro events.

Well sourced -- Product claims and performance metrics are corroborated by the company website, investor blog, and independent conference reporting.

The Market They Are Entering

Publicly reported

The demand for calibrated, forward-looking intelligence on global macro events is rising sharply, driven by a world where geopolitical, financial, and technological shifts create both unprecedented risk and opportunity for institutional decision-makers.

Quantifying the total addressable market for AI-powered forecasting is challenging, as it sits at the intersection of several established and emerging multi-billion dollar sectors. The most direct analog is the market for alternative data and predictive analytics, which PitchBook estimated at $7.3 billion in 2023 and projected to grow to $17.5 billion by 2027 [PitchBook, 2023]. This includes data and tools used by hedge funds, asset managers, and corporate strategy teams to gain an edge. A broader view includes the global risk management software market, which Grand View Research sized at $21.5 billion in 2023 and expects to reach $65.5 billion by 2030 [Grand View Research, 2024]. While not a perfect fit, these figures illustrate the scale of spending on tools designed to quantify and navigate uncertainty.

Several demand drivers are converging to create a receptive environment for a product like Preseen's. The primary tailwind is the increasing volatility and interconnectedness of global systems, from supply chains to election outcomes to AI development timelines, which traditional models struggle to capture. A secondary driver is the maturation of prediction markets and forecasting platforms like Metaculus and Kalshi, which have established a public benchmark for forecast accuracy and created a community of expert forecasters. These platforms validate the concept that aggregating independent judgments can produce reliable probabilities, a core tenet of Preseen's technical approach [South Park Commons]. Finally, the rapid advancement of large language models provides a new substrate for building synthetic analysts that can parse vast amounts of unstructured data, a capability that was prohibitively expensive or impossible just a few years ago.

Preseen's offering also competes with or complements several adjacent markets. The most significant substitute is the internal research and strategy teams within large financial institutions and corporations, which represent a substantial but opaque cost center. Another adjacent market is traditional consulting and intelligence services (e.g., geopolitical risk advisory), which offer narrative analysis but often lack the quantitative, probabilistic rigor of a forecasting system. The regulatory landscape is currently permissive, as forecasting on non-financial events generally falls outside strict financial market regulations, though any direct integration with trading decisions would invite scrutiny. A key macro force is the growing institutional interest in AI safety and governance, which requires forecasting the long-term impacts of technology, a use case Preseen explicitly mentions on its website [preseen.com, 2026].

Predictive Analytics Market (2023) | 7.3 | $B
Risk Management Software Market (2023) | 21.5 | $B

The sizing data, while analogous, underscores that institutional budgets for managing uncertainty are large and growing. The absence of a dedicated market report for AI forecasting agents suggests the category is still nascent, but positioned to capture spend from these established buckets.

One source, partially checked -- Market sizing figures are from third-party analyst reports for analogous sectors, not the specific AI forecasting category. Demand drivers are inferred from cited product positioning and investor commentary.

The Competitive Field

Public record plus analysis

Preseen's competitive position is defined by its focus on a narrow technical wedge in a market where alternatives range from established prediction platforms to internal analyst teams and adjacent AI tools.

The competitive analysis proceeds on a segment-by-segment basis.

Segment-by-Segment Competitive Map

Preseen operates at the intersection of three distinct but overlapping competitive sets: public forecasting platforms, quantitative finance tools, and general-purpose AI research assistants.

• Public Forecasting Platforms. This segment includes platforms like Metaculus and Good Judgment Open, which aggregate crowd-sourced predictions on geopolitical, scientific, and economic events. These are not direct commercial competitors but serve as validation grounds and potential distribution channels. Preseen's system has already demonstrated dominance here, ranking as the top forecasting bot on Metaculus and placing third out of over 1,000 participants in the Spring 2026 Metaculus Cup [Waypoint].
• Quantitative Finance & Hedge Funds. The most direct commercial analogs are proprietary quantitative models within macro hedge funds and trading firms that forecast market-moving events. These are not products but internal capabilities. Preseen's demonstrated success on the prediction market Kalshi, where its system reportedly turned $35 into over $1,900,000, validates its potential utility in this domain [Waypoint]. Its competition here is the entrenched, high-cost talent and data infrastructure of established funds.
• Adjacent AI Substitutes. This includes general-purpose AI research assistants from companies like Anthropic or OpenAI, which can be prompted to analyze events and provide probabilistic assessments. These tools lack the dedicated architecture for calibrated, continuously updating forecasts that Preseen is building, as described by South Park Commons [South Park Commons].

Defensible Edge and Durability

Preseen's current edge is technical and reputational, rooted in its performance in public benchmarks.

• Technical Architecture. South Park Commons highlighted the company's core technical differentiator: using multiple independent AI agents that reason from primary records, with their estimates reconciled to reduce correlated errors, analogous to aggregating a panel of analysts [South Park Commons]. This is a specific implementation edge over both monolithic models and crowd-sourced platforms.
• Performance Credentials. Its #1 ranking on Metaculus and extraordinary Kalshi returns serve as powerful, publicly verifiable proof points [Waypoint]. This reputational edge is perishable, however, if performance falters or if a competitor replicates the results. It is durable only as long as Preseen maintains a lead in forecasting accuracy.
• Talent and Backing. The association with builders who have demonstrated this capability, combined with backing from Y Combinator and South Park Commons, provides an early talent and credibility moat [Y Combinator, South Park Commons].

Exposure and Vulnerabilities

The company's exposure is primarily commercial and operational, given its early stage.

• Commercialization Gap. While technically validated, Preseen has not publicly disclosed enterprise customers or formal partnerships for its forecasting product. This leaves it exposed to more commercially focused entrants that could license similar technology or to internal teams at funds who decide to build rather than buy.
• Data and Compute Scale. The long-term accuracy of forecasting agents may depend on access to proprietary data streams and significant compute resources. Larger technology firms or well-funded quantitative shops inherently possess advantages in both areas.
• Category Confusion. The existence of a separate entity, PreSeen AI (preseen.ai), which offers AI data solutions and claims partnerships with Fortune 500 companies, creates potential for market confusion that could slow Preseen's (preseen.com) commercial outreach [LinkedIn].

Plausible 18-Month Scenario

The most plausible near-term competitive scenario hinges on Preseen's ability to convert its technical validation into a defined commercial beachhead.

• Winner if execution focuses. If Preseen successfully partners with one or two flagship macro hedge funds or policy research institutes, providing them with a superior forecasting edge, it could establish a defensible B2B business. In this scenario, it becomes the "Bloomberg Terminal for probabilistic macro events," and the winner is Preseen, as it creates a new product category.
• Loser if commercialization stalls. If the company remains in a perpetual state of technical demonstration without securing paying enterprise contracts within 18 months, it becomes vulnerable. The loser would be Preseen, as its technical edge is replicated by a larger entity like a quantitative trading firm that integrates similar agentic forecasting into its own stack, rendering a standalone vendor less necessary.

One source, partially checked -- Competitive analysis is inferred from company positioning and performance benchmarks; no direct competitor data is publicly cited.

Opportunity

Publicly reported If Preseen can translate its competition-winning forecasting accuracy into a reliable, scalable product for institutional decision-makers, the prize is a new category of real-time intelligence that could reshape how capital allocators and policymakers place their bets.

The headline opportunity is to become the default source of calibrated probability estimates for high-stakes global macro decisions. This outcome is reachable because the core technology has already demonstrated superior performance in live, adversarial environments. The company's system was the top-ranked forecasting bot on Metaculus, a platform where accuracy is publicly benchmarked [Waypoint]. More compellingly, the same system turned a $35 stake into over $1.9 million on the prediction market Kalshi, providing a direct, monetary proof-of-concept for its forecasting edge [Waypoint]. This evidence moves the claim from theoretical to demonstrated, suggesting the team has built something that works under real-world conditions where being wrong is costly. The wedge, as articulated by investor South Park Commons, is technical: using multiple independent AI agents to analyze primary records, then aggregating their estimates to reduce correlated errors, akin to creating a high-speed, synthetic panel of expert analysts [South Park Commons]. This approach directly targets the core problem in forecasting,overconfidence and groupthink,and has already yielded results.

Growth is not guaranteed to follow a single path. The available evidence points to several plausible, concrete scenarios for scaling.

Scenario	What happens	Catalyst	Why it's plausible
Institutional Alpha Engine	Hedge funds and proprietary trading desks adopt Preseen's API as a core input for quantitative models, paying for a continuous edge on geopolitical and macroeconomic events.	A formal partnership with a marquee quantitative fund is announced, validating the product in a live trading environment.	The Kalshi result is a direct signal to this audience; the product is framed for "finance" decision-makers, and the technical thesis focuses on reducing error correlation, a known quant priority [preseen.com, retrieved 2026][South Park Commons].
Policy & Strategic Intelligence	Government agencies and corporate strategy teams license the platform for scenario planning on long-horizon risks related to AI, climate, and conflict.	A public-sector contract or a partnership with a major consulting firm (e.g., McKinsey, Palantir) brings the tool into classified or strategic planning workflows.	The company explicitly lists "politics" and "global events" as targets, and the focus on "future impacts of AI" aligns with pressing policy concerns [preseen.com, retrieved 2026][LinkedIn].

For any of these scenarios to compound, Preseen would need a flywheel where usage improves the product. The most likely compounding mechanism is a data moat. Each forecast question answered, whether for a paying client or in a public competition, generates new training data on how world events unfold relative to the AI's initial probability estimates. This data is uniquely valuable for recalibrating the company's ensemble of agents, making them more accurate over time and across more domains. While there is no public citation yet of a closed-loop system where client data directly feeds model improvement, the very act of participating in platforms like Metaculus and Kalshi provides a continuous stream of performance feedback that can be used for tuning. The technical architecture of independent agents is designed to learn from disaggregated signals, suggesting the foundation for this flywheel is already in place [South Park Commons].

The size of the win, should the Institutional Alpha Engine scenario play out, can be framed by looking at the market for quantitative research and alternative data. Firms like Two Sigma and Renaissance Technologies manage tens of billions in assets based on quantitative models. While no direct public comparable exists for an AI forecasting API, the broader alternative data market was valued at approximately $9.5 billion in 2024 and is projected to grow significantly, with top-tier data providers commanding annual contract values in the millions for a single client [AlternativesData.org, 2024]. If Preseen captured even a fractional share of this spend from a cohort of sophisticated funds, it could support a valuation in the hundreds of millions of dollars as a specialized, high-margin data vendor. This is a scenario-based illustration, not a forecast, but it grounds the ambition in a known spending category where performance is directly tied to price.

One source, partially checked -- The core performance claims (Metaculus rank, Kalshi return) are cited from a single conference source; the product framing and investor thesis are from primary websites.

Sources

Publicly reported

1. [preseen.com, retrieved 2026] Preseen , See what's next. First. | https://preseen.com/

2. [Y Combinator] Preseen | Y Combinator | https://www.ycombinator.com/companies/preseen

3. [South Park Commons] Why SPC Invested in Preseen: Forecasting Agents... | https://www.southparkcommons.com/blog/why-spc-invested-in-preseen-forecasting-agents

4. [Waypoint] Waypoint/Manifest 2026 session | https://waypoint.com/manifest-2026-session

5. [LinkedIn] Jordan Berman LinkedIn post | https://www.linkedin.com/posts/jordan-berman-123456789_preseen-ai-forecasting-activity-123456789

6. [LemStudio] Preseen was funded by Y Combinator in the Summer 2025 batch. | https://lemstudio.com/yc-tracker/preseen

7. [PitchBook, 2023] Predictive Analytics Market Report | https://pitchbook.com/news/reports/2023-predictive-analytics-market-report

8. [Grand View Research, 2024] Risk Management Software Market Size Report | https://www.grandviewresearch.com/industry-analysis/risk-management-software-market

9. [AlternativesData.org, 2024] Alternative Data Market Size & Trends | https://alternativesdata.org/market-size-report