The Form 13F-HR filing of Numerai GP LLC, dated September 30, 2025, has been published.

You can use my short program analysis 12 quarters of 13F-HR (reporting date 2022/12/31 to 2025/09/30) in the Excel file. For a deeper dive into the 11-quarter trend, refer to the Excel dataset and Jupyter notebook analysis available on GitHub.


History 13F-HR in the Excel file


Analysis_Numerai_GP.ipynb


The value of long positions (AUM from 13F)

To reiterate, only long positions are reported on Form 13F-HR. Short positions are not reported.


     Date         Long positions (billion dollars)
2022-12-31 $0.15
2023-03-31 $0.30
2023-06-30 $0.49
2023-09-30 $0.30
2023-12-31 $0.28
2024-03-31 $0.23
2024-06-30 $0.28
2024-09-30 $0.38
2024-12-31 $0.37
2025-03-31 $0.53
2025-06-30 $0.69 
2025-09-30 $0.71

This quarter’s long positions amount to $0.71 billion. The long positions of $0.71 billion represent the largest value reported in past 13F-HR filings, but it is not an outlier.

Number of holdings

In the current quarter, the number of holdings has resumed its upward trend and continues to increase.


Date HOLDINGS
2022-12 111
2023-03 156
2023-06 159
2023-09 219
2023-12 232
2024-03 250
2024-06 311
2024-09 345
2024-12 389
2025-03 490
2025-06 438
2025-09 546

Position Size

Portfolio diversification seems to have progressed further, which becomes especially evident when compared to the last quarter. This can be seen in the following chart, which shows an increase in the number of smaller positions.

The Numerai team appears to be continuing to build a portfolio defined by diversified holdings and a larger number of smaller positions.




In Conclusion

Although the 13F-HR report is limited to U.S. market securities and only discloses long positions, the following insights can still be inferred.

  • JPMorgan Securities’ investment did not immediately affect the value of Numerai hedge fund’s long positions.
            August 27, 2025  https://x.com/richardcraib/status/1960358823576928497

  • The correlation between the value of long positions and holdings has increased since 2024.

Since 2022, the correlation between long positions and holdings has stood at 0.77, but since 2024 it has risen to 0.90. The initiative to continue building a portfolio characterized by diversified holdings and numerous smaller positions seems to have remained effective since 2024. It seems that the optimizer, along with updates to data changes and staking rules, is working effectively.

Reference: Numerai as Decentralized AI: Insights from 13F-HR and Operational Costs, p. 6.

Extra: Having a Generative AI Break Down a Portfolio Built by Decentralized AI

I had Gork analyze the History 13F-HR, which I converted into a CSV file from the Excel file.

Prompt

-This CSV file summarizes hedge fund 13F-HR filings.
-The columns are as follows:
･nameOfIssuer: Company name of the Long Position
･cusip: CUSIP of the Long Position
･From "2022/12/31" to "2025/9/30": Amount of the Long Position
-The rows contain the following:
For each nameOfIssuer and cusip, the Long Position amounts from "2022/12/31" to "2025/9/30"

-Please generate a comprehensive report covering:
･Portfolio trends by sector (e.g., Tech, Healthcare, Financials)
･Strategic shifts over time and correlation with macroeconomic factors (e.g., interest rate hikes, geopolitical risks)
･Fund characteristics: growth vs. value focus, active vs. passive style, and thematic concentration (e.g., AI, renewable energy)
･Risk management: diversification level and defensive stock ratio
･Forward-looking insights: forecast future portfolio directions based on historical trends, sector rotation patterns, thematic focus, and macro sensitivity

What Gork Said

Note: Grok realized it was Numerai’s 13F-HR because the file name contained the keyword "Numerai."

Comprehensive Report on Numerai's 13F-HR Portfolio (Q4 2022 - Q3 2025)

Executive Summary
Numerai's portfolio, as detailed in the provided 13F-HR filings CSV, demonstrates robust growth in assets under management (AUM) from $147 million in Q4 2022 to $709 million in Q3 2025, achieving a compound annual growth rate (CAGR) of approximately 70%. The number of positions expanded from 111 to 546, reflecting increased diversification. The fund maintains a growth-focused, active management style with thematic emphases on AI-driven technology and renewable energy. Sector trends indicate early dominance by technology and energy, a shift to financials, healthcare, and consumer staples during 2023's rate-hike environment, and a recent pivot to industrials amid easing monetary policy. Strategic adjustments correlate with macroeconomic factors, including Federal Reserve interest rate hikes (peaking at 5.25-5.50% in 2023), geopolitical tensions (e.g., Russia-Ukraine conflict driving energy volatility), and oil price fluctuations (WTI from highs of ~$120/barrel in 2022 to ~$60/barrel in 2025). Risk management has strengthened through broader diversification, with defensive stocks comprising 20-30% of holdings. Forward insights point to sustained industrials and AI focus, potential rotations to defensives if recession signals emerge, and sensitivity to ongoing rate cuts and geopolitical risks.

Portfolio Trends by Sector
Sector allocations were derived by categorizing the 1,408 unique issuers based on GICS standards, using top holdings (which represent ~25-35% of AUM per quarter) as proxies for trends and cross-referencing with known classifications (e.g., Zeta Global as Technology, Teekay Tankers as Energy). Total AUM grew unevenly, peaking at $694 million in Q2 2025 before a slight uptick. Allocations emphasize growth sectors, with rotations evident.
  • Technology (average 30-40%): Strong early (Q4 2022: ~35%, led by Zeta Global at $5.2M, Semtech, Upwork) but moderated to ~25% in 2023 amid rate pressures. Rebounded to ~40% by Q1 2025 (e.g., Arista Networks at $4.9M, Salesforce, Credo Technology), driven by AI and software (HubSpot, Datadog). By Q3 2025, ~35% with Impinj and Snowflake.
  • Industrials (15-25%): Gained prominence from Q3 2024 (~20%), becoming dominant by Q3 2025 (~30%, e.g., IDEX Corp at $10.1M, Hubbell, Heico). Early exposure limited (~10%, e.g., L3Harris), but infrastructure themes (Valmont, Myr Group) fueled growth.
  • Financials (10-20%): Peaked in Q1 2023 (~25%, e.g., Zions at $9.4M) as rates rose, benefiting banks and insurers (Everest Group, Manulife at $10.2M in Q3 2025). Stabilized at 15-20%, including fintech (SoFi, PayPal).
  • Healthcare (15-20%): Defensive tilt in 2023 (~25%, Baxter at $9.4M, Illumina), dipping mid-period before rebounding (~20% in 2025, Procept Biorobotics, Tempus AI). Focus on biotech and medtech.
  • Consumer Discretionary (10-15%): Volatile, high in Q2 2023 (~20%, DraftKings at $10.5M, Roblox) and Q2 2025 (Wyndham at $10.6M). Includes retail (Costco at $11.8M in Q3 2025) and leisure.
  • Energy (5-15%): Elevated early (Q4 2022: ~15%, Teekay at $5.2M, Nordic American) due to geopolitical-driven price spikes, declining to ~5% by 2025 but with renewables (Enphase, Bloom Energy) and traditional (Expand Energy at $8.9M).
  • Materials (5-10%): Intermittent (Q2 2023: ~15%, Nutrien at $9.6M, Alcoa), tied to commodities. Recent mining focus (New Gold, Franco-Nevada).
  • Consumer Staples (5-10%): Rose defensively in 2023 (Tyson at $9.2M), stable at ~10% (Celsius, Hormel).
  • Other (Utilities, Communication: <10%): Minor; utilities like NRG peaked at $6.8M in Q4 2023.
QuarterTech (%)Industrials (%)Financials (%)Healthcare (%)Consumer Disc. (%)Energy (%)Materials (%)Staples (%)Other (%)
Q4 2022351010151015555
Q1 202330102520555105
Q2 2023251515152051555
Q3 202330201010155555
Q4 2023251510201010555
Q1 202430151015105555
Q2 202435201010105555
Q3 202435251010105555
Q4 202430201510155555
Q1 202525301510105555
Q2 20253025155155555
Q3 202525302010105555
Trends show rotation from volatile growth (tech/energy) to value/defensives (financials/healthcare) in 2023, reverting to growth (tech/industrials) post-2024.
Strategic Shifts Over Time and Correlation with Macroeconomic Factors

Portfolio evolution occurred in phases, aligning with macro dynamics:

  1. Q4 2022 - Q2 2023 (Growth to Value): Tech and energy heavy initially (e.g., Zeta, Teekay), shifting to financials (Zions) and defensives (Baxter, Tyson). Correlated with Fed rate hikes (from 0.25% in Mar 2022 to 5% by May 2023), compressing growth valuations, and Russia-Ukraine invasion (Feb 2022) spiking WTI oil to $120/barrel in Mar 2022, boosting shipping/energy.
  2. Q3 2023 - Q4 2024 (Diversification and Recovery): Positions doubled; consumer/materials rose (DraftKings, Nutrien) as inflation eased but recession fears persisted. AUM dipped in Q4 2023 amid volatility. Correlated with rate pause at 5.25-5.50% (Jul 2023) and Middle East tensions (e.g., Israel-Hamas from Oct 2023), stabilizing oil ~$70-90/barrel.
  3. Q1-Q3 2025 (Growth Rebound): Tech and industrials surged (Arista, IDEX) with rate cuts (to ~4% by mid-2025), favoring growth. Oil dipped to ~$60/barrel amid reduced tensions, shifting energy to renewables.
Shifts reflect macro sensitivity: rate hikes favored value, cuts revived growth; geopolitics (Ukraine, Middle East, US elections 2024) influenced energy/commodities.

Fund Characteristics

  • Growth vs. Value Focus: Primarily growth-oriented (60-70%), with high-beta holdings (e.g., DraftKings, Roblox). Value emerged in 2023 (banks, staples) but overall emphasizes high-P/E AI/tech.
  • Active vs. Passive Style: Active, with ~50% quarterly turnover in top holdings (e.g., Zeta to DraftKings to IDEX). No index mirroring; stock-picking evident in thematic selections.
  • Thematic Concentration: AI/digital (15-20%, Zeta, Arista, Salesforce); renewables (Enphase, Bloom); fintech/disruption (Upwork, SoFi). Aligns with Numerai's AI roots.

Risk Management

  • Diversification Level: Enhanced; Herfindahl-Hirschman Index fell from ~0.12 (concentrated) in 2022 to ~0.04 (diversified) in 2025. Top 10 holdings ~25-35% of AUM.
  • Defensive Stock Ratio: 20-30%, higher in volatile periods (e.g., 35% in 2023 with healthcare/staples). Beta ~1.1-1.3; balanced volatile tech/energy with staples/utilities.
Forward-Looking Insights

Historical trends (AUM growth, 6-12 month rotations) and themes (AI, renewables) suggest continued industrials emphasis if economy strengthens (e.g., GDP >3%), with expansions in infrastructure (IDEX-like). AI/tech could reach 40% amid adoption, especially with rates potentially below 4% by 2026. If recession risks rise (e.g., from geopolitics like escalating Middle East conflicts), rotate to staples/utilities (Costco scaling). Energy sensitivity to oil (~$68 average 2025) may prompt renewables focus. AUM could exceed $1B by 2026; monitor Fed policy and elections for shifts.


My Thoughts

  • Gork may be better at processing CSV files than Excel files.
  • In the table shown under “Portfolio Trends by Sector,” the totals of each row exceed 100%. Gork acknowledged this issue as well.
When a prompt is entered and the data is registered into the LLM, a report is generated within a few minutes.
However, checking whether the report created in just a few minutes is correct would probably take me several hours. Moreover, since I am not a financial expert, even if I spend time reviewing the report, I might still miss some mistakes.
It seems that LLMs still make errors, but the kinds of mistakes I can notice eventually will disappear in the future.
On the other hand, what should be done about the mistakes I cannot detect?
Should I not use LLMs outside my own area of expertise? Is it more appropriating for me to use them within my specialty to improve efficiency?
In the future, will those who can point out the mistakes made by LLMs be recognized as experts?