Funds & ETFsFlows, fees & vehicles
Ex-poker pro David Orr describes using AI for Japanese filings and risk monitoring, and warns AI compute could go through a boom and bust
On Excess Returns, Militia Capital's David Orr describes using AI to translate Japanese filings, rank news and track exposures shared across positions. He argues AI compute may follow a commodity-style boom and bust, so using AI well differs from investing in it.
From the Excess Returns podcast episode of November 17, 2026 and cited sources: Orr presents AI as research support rather than autonomous trading. [1] [4]
SOURCES[1] From Poker Pro to Elite Hedge Fund Manager: David Orr on… podcasters.spotify.com[4] HedgeFundAlpha: Militia Capital Full Year and Q4 2023… hedgefundalpha.com
AI-assisted, not human-edited. Methodology Corrections Who owns this site
- Original source
- Excess Returns podcast episode, independent perspective
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- When
- November 17, 2026 podcast episode published
- Covered
- Nov 18 edition
- Evidence
- 11 sources across 11 sites; no transcript
- 1From Poker Pro to Elite Hedge Fund Manager: David Orr on…podcasters.spotify.com
- 2StockAnalysis: Militia Long/Short Equity ETF (ORR)stockanalysis.com
- 3ORR Prospectus (ETF Architect)etfarchitect.com
- 4HedgeFundAlpha: Militia Capital Full Year and Q4 2023…hedgefundalpha.com
- 5AUM13F: Militia Capital Management LLCaum13f.com
- 6Dakota: Militia Capital Management, LLCdakota.com
Reach: not measured yet
Orr presents AI as research support rather than autonomous trading. He uses it to translate Japanese disclosures, decide which news matters most and combine risk shared across hundreds of small positions. Japan is central to the work: in 2023 about a quarter of the long book was in Japanese stocks. No model, vendor, prompts or error rates are named. [1] [4]
Orr expects AI compute to go through a commodity-style boom and bust. He also separates benefiting from AI from owning AI infrastructure. A separate interview summary links him to the view that intelligence will become abundant while the physical world lags behind. No reviewed source tests the forecast. [1] [7]
a seven-person firm uses AI for translation, news triage and risk aggregation across hundreds of positions.
Read the full assessment
Implication: these are low-risk starting points for small teams, but the gains are unmeasured and the compute-cycle thesis is one practitioner's view.
Sources are gathered (the fund's filings and prospectus, ETF data, the 2023 investor letter and third-party summaries). Writing the dossier now.
Executive brief
David Orr spent a decade as a professional poker player. Eight days before this episode aired, his firm's ETF had about $363M in assets, and the fund's prospectus lists total annual expenses of 10.91%. Most of that, 9.61%, is the cost of carrying short positions (StockAnalysis; ORR prospectus). On Excess Returns he says he uses AI to translate Japanese filings, rank news and monitor portfolio risk. He also expects AI compute to go through a commodity-style boom and bust. He argues that using AI well is a different skill from picking which AI companies will reward shareholders. This is commentary from a practitioner, not a technical disclosure.
What changed and event timeline
2023 letter shows a big long/short gap
Militia reported a 2023 gross return of +15.3%, against +26.2% for the S&P 500. Its longs gained 58% and its shorts fell 27%. Assets were $58M, spread across about 600 companies ().
ORR ETF launches
The Militia Long/Short Equity ETF, an actively managed global long/short fund, began trading with Orr as portfolio manager (;).
Regulatory assets reach $958.6M
A 13F aggregator lists Militia Capital Management's regulatory assets at that figure, based in Austin with seven employees. That is far above the 2023 total (;).
Excess Returns episode airs
Orr and Jack Forehand discuss AI in research, the case for a compute oversupply, crowded shorts and Japan ().
ORR closes at $39.52
The fund had about $363M in assets and 278 holdings, with a one-year return of 19.9% ().
Capabilities and access
- AI tools: neither the episode description nor any other reviewed source names a model, vendor or version.
- Uses described (episode, ch. 15:53):
- translating Japanese filings
Read the full section
- AI tools: neither the episode description nor any other reviewed source names a model, vendor or version.
- Uses described (episode, ch. 15:53):
- translating Japanese filings
- prioritizing news
- monitoring exposures shared across positions
- Investor access: retail investors can buy the strategy through the Nasdaq-listed ORR ETF. The hedge fund, Militia Capital Partners LP, is a private fund (Dakota).
Technical analysis for researchers and developers
The described workflow is AI support for research, not autonomous trading. No prompts, evaluations or error rates are documented.
Read the full section
The described workflow is AI support for research, not autonomous trading.
- Language arbitrage. Machine translation of Japanese disclosures lets a small team cover many companies. About 25% of the fund's long book was already in Japanese stocks in 2023 (HedgeFundAlpha).
- Shared-exposure monitoring. With hundreds of small positions, the risk sits in factors the positions have in common, so AI is aimed at aggregation rather than picking stocks.
- Consensus bias. Orr says AI models tend to repeat consensus views, which leaves room for human judgment (episode, ch. 20:58).
No prompts, evaluations or error rates are documented.
Claims and evidence
- Compute may turn into a boom and bust: this is Orr's opinion. No independent analysis supports or tests the forecast.
- Poker background: he played roughly 10 million hands and ranked in the global top 50 in No-Limit Hold'em.
- Performance: fund letters report it as gross returns. ORR's NAV return is third-party market data (StockAnalysis).
Read the full section
- Compute may turn into a boom and bust: this is Orr's opinion. A third-party summary of another Orr interview attributes to him the view that intelligence will become abundant while the physical world lags (Podwise). No independent analysis supports or tests the forecast.
- Poker background: he played roughly 10 million hands and ranked in the global top 50 in No-Limit Hold'em. These are self-reported figures that appear in fund materials (search summary of Militia materials).
- Performance: fund letters report it as gross returns. ORR's NAV return is third-party market data (StockAnalysis).
Context and prior work
- Orr's case for an oversupply resembles long-running "capital cycle" arguments discussed elsewhere on Excess Returns (Edward Chancellor episode).
- It also resembles efficiency-driven arguments about GPU demand (Trask).
- He has described his poker-to-hedge-fund path before, on another podcast (Odds On Open).
Limitations, safety and contested findings
- Shorting is costly. The 9.61% in short-related expenses inside ORR's 10.91% total shows what the strategy pays to carry its shorts (prospectus).
- Despite the fund's many positions, its largest holding, Hikari Tsushin, was about 12% of assets (StockAnalysis).
- 13F data is partial. The aggregator's reported holdings and dates are uneven, so the 13F gives only a partial view (AUM13F).
Read the full section
- Shorting is costly. The 9.61% in short-related expenses inside ORR's 10.91% total shows what the strategy pays to carry its shorts (prospectus).
- Concentration in the ETF. Despite the fund's many positions, its largest holding, Hikari Tsushin, was about 12% of assets (StockAnalysis).
- 13F data is partial. The aggregator's reported holdings and dates are uneven, so the 13F gives only a partial view (AUM13F).
- Orr's own caveats. The episode itself warns about crowded shorts and leverage-driven selloffs, and says index funds suit many individual investors.
Business and practitioner implications
- The firm reports seven staff covering global small caps (AUM13F).
- Orr separates benefiting from AI from owning AI infrastructure.
- Market facts, not advice. Every figure here reports what happened. None is a recommendation.
Read the full section
- AI as leverage for small teams. The firm reports seven staff covering global small caps (AUM13F). Translation and risk monitoring are useful, low-risk places to start with AI, but the reviewed sources give no measured productivity gains.
- Users versus investors. Orr separates benefiting from AI from owning AI infrastructure. His concern about a compute glut is an attributed opinion, not a forecast backed by data.
- Market facts, not advice. Every figure here reports what happened. None is a recommendation.
The source trail.
Sources (11)
From Poker Pro to Elite Hedge Fund Manager: David Orr on Investing, Risk and AI
Metadata only. Transcript unavailable: Audio transcription returned HTTP 429.
podcasters.spotify.com