Human judgment × machine intelligence
DTRM Research Lab tests a precise proposition: in uncertain, open systems, computational power creates more value when human judgment retains control of the objective, constraints and final decision.
Narrative protocol
Six connected chapters trace the argument from bounded games to a second real-world decision domain, without claiming victories the research has not yet demonstrated.
DeepBlue, AlphaGo and AlphaStar demonstrated extraordinary search and optimization inside bounded environments. Their victories begin after humans have already specified what the game is and what winning means.
The machine solves the game. The human defines it.
A normalized comparison between human-controlled selection, robust machine portfolio construction, machine recommendations, AI-managed instruments and the S&P 500 benchmark.
The experiment treats the market as a repeated, adaptive decision environment. The model searches and ranks before the decision; portfolio construction is the action; realized return and risk are observations after the action.
Trump Score and Adjusted Score are ex-ante signals. News evidence confidence shrinks the underlying prediction before those scores are formed. Calmar, Sharpe, volatility and maximum drawdown are ex-post diagnostics and never leak back into the frozen decision.
The canonical backend risk report remains defined for DTRM Fund, ChatGPT Fund, AIEQ and the S&P 500. For display parity, MM1 cumulative return, Calmar and maximum drawdown use the shared-inception comparison path: the published DTRM path through the 2026-09-25 close, followed only by prospective MM1 observations. This continuation does not invent pre-fork MM1 history.
MM1 and DTRM now share the same eleven-player 1-4-4-2 decision geometry. What changes is the selector: human judgment over the evidence-aware ranking versus direct robust portfolio optimization.
Read the cohort amendment →| Rank | Movement | Asset | Adjusted | Trump score | Signal |
|---|---|---|---|---|---|
| 01 | VVisa Inc. | Adjusted score: 83.3 | Trump Score: 83.9 | strong | |
| 02 | MAMastercard Incorporated | Adjusted score: 79.5 | Trump Score: 81.6 | strong | |
| 03 | SOSO | Adjusted score: 75.9 | Trump Score: 83.3 | strong | |
| 04 | TMOTMO | Adjusted score: 75.2 | Trump Score: 79.3 | strong | |
| 05 | HCAHCA | Adjusted score: 74.8 | Trump Score: 85.0 | strong | |
| 06 | TTTT | Adjusted score: 74.6 | Trump Score: 82.2 | strong | |
| 07 | WMWM | Adjusted score: 73.3 | Trump Score: 82.7 | strong | |
| 08 | MCKMCK | Adjusted score: 71.1 | Trump Score: 76.4 | strong | |
| 09 | MARMAR | Adjusted score: 71.1 | Trump Score: 80.4 | watch | |
| 10 | RYRY | Adjusted score: 70.7 | Trump Score: 75.3 | watch | |
| 11 | SBUXSBUX | Adjusted score: 70.3 | Trump Score: 79.9 | watch | |
| 12 | GILDGILD | Adjusted score: 69.6 | Trump Score: 72.4 | watch | |
| 13 | AXPAmerican Express Company | Adjusted score: 69.6 | Trump Score: 78.7 | watch | |
| 14 | DHRDHR | Adjusted score: 68.9 | Trump Score: 81.0 | watch | |
| 15 | UNHUnitedHealth Group Incorporated | Adjusted score: 68.4 | Trump Score: 78.1 | watch | |
| 16 | VRTXVRTX | Adjusted score: 67.9 | Trump Score: 75.8 | watch | |
| 17 | ELVELV | Adjusted score: 67.7 | Trump Score: 77.0 | watch | |
| 18 | CVSCVS | Adjusted score: 66.4 | Trump Score: 84.5 | defensive | |
| 19 | CBCB | Adjusted score: 65.5 | Trump Score: 68.9 | defensive | |
| 20 | JPMJPMorgan Chase & Co. | Adjusted score: 65.1 | Trump Score: 67.2 | defensive | |
| 21 | MCDMcDonald's Corporation | Adjusted score: 64.6 | Trump Score: 72.9 | defensive | |
| 22 | UBERUBER | Adjusted score: 64.5 | Trump Score: 77.6 | defensive | |
| 23 | COFCOF | Adjusted score: 63.1 | Trump Score: 71.2 | defensive | |
| 24 | ADPADP | Adjusted score: 63.0 | Trump Score: 68.3 | defensive | |
| 25 | HONHON | Adjusted score: 62.7 | Trump Score: 74.7 | defensive | |
| 26 | AZNAstraZeneca PLC | Adjusted score: 62.2 | Trump Score: 70.1 | defensive | |
| 27 | AMGNAMGN | Adjusted score: 61.7 | Trump Score: 73.5 | defensive | |
| 28 | ABBVAbbVie Inc. | Adjusted score: 60.9 | Trump Score: 62.6 | defensive | |
| 29 | PGRPGR | Adjusted score: 60.0 | Trump Score: 69.5 | defensive | |
| 30 | JCIJCI | Adjusted score: 59.5 | Trump Score: 65.5 | defensive |
DTRM and MM1 share the same eleven-player 1-4-4-2 geometry. The selection mechanism is the experimental variable. ChatGPT remains the independent memory-free advisory baseline.
The production ranking applies news-evidence confidence before Trump Score and Adjusted Score are formed. The human then selects the final eleven from that governed search surface and assigns the fixed football-role weights.
MM1 optimizes the actual eleven-player portfolio rather than a Top-25 intermediate set. It uses the evidence-adjusted nominal prediction and calibrated P10 downside state under the same 1-4-4-2 weight geometry, with the robust uncertainty budget frozen before the cohort begins.
ChatGPT independently selects eleven assets and proportional weights under the recurring memory-free advisory protocol. It does not receive the DTRM or MM1 portfolio construction rule as its decision mechanism.
One private daily batch curates the same FMP evidence set used by the model. DTRM agent claims describe possible impact on published positions; they never alter the locked 60-day alignments.
15 Oct, 14:29 UTCbusinessinsider.com (opens in a new tab)
Anthropic allowing external AI-safety evaluators could increase scrutiny and compliance emphasis across frontier AI development, which may modestly raise perceived regulatory/operational risk for big AI platforms and infrastructure names held (GOOGL, META, MSFT, NVDA, AVGO).
15 Oct, 13:43 UTCreuters.com (opens in a new tab)
Amazon’s launch of two portable AI listening devices could signal incremental device-led AI demand that may be supportive for AI compute and platform ecosystems relevant to holdings (NVDA, AVGO, GOOGL, META, MSFT), though the direct linkage is unclear from the article text.
15 Oct, 16:11 UTCFinancial Modeling Prep (opens in a new tab)
The article’s emphasis on strong AI-driven enterprise demand (via Accenture commentary) could be read as supportive for continued AI spending, which may benefit AI infrastructure and platforms in the portfolio (NVDA, AVGO, MSFT, GOOGL, META).
15 Oct, 11:21 UTCmarketwatch.com (opens in a new tab)
A reported continued selloff in the AI/tech trade and rotation into bonds/defensives could pressure the fund’s large-cap AI/tech holdings (NVDA, AVGO, GOOGL, META, MSFT) in the near term, while potentially being less negative or relatively supportive for MMM as a more defensive industrial name.
15 Oct, 14:19 UTCetftrends.com (opens in a new tab)
Discussion of improving access to bitcoin via ETFs, stablecoins, and tokenization could imply gradual growth in related transaction and banking activity, which may be modestly supportive for payments and bank holdings (V, MA, JPM, BAC, RY).
15 Oct, 14:30 UTCzacks.com (opens in a new tab)
Vicor’s raised revenue outlook attributed to surging AI demand could reinforce the narrative of strong AI data-center buildout, which may be supportive for AI compute and connectivity suppliers held (NVDA, AVGO) and related platforms (MSFT, GOOGL, META).
15 Oct, 14:30 UTCseekingalpha.com (opens in a new tab)
An analyst upgrade citing rapidly expanding AI data-center and optical-module demand could be read as supportive of continued data-center capex trends, which may benefit portfolio AI infrastructure names (NVDA, AVGO) and major AI platforms (MSFT, GOOGL, META).
15 Oct, 14:24 UTC247wallst.com (opens in a new tab)
The article attributes META’s recent outperformance to strength of its newest AI model, which could be supportive for the portfolio’s META position even as it highlights potential volatility around whether gains are sustainable.
LLM-assisted research annotation · Probabilistic claims · Not financial advice
Checked-in fallback evidence plus the frozen prospective MM1 cohort decision. Live MM1 observations replace the bootstrap anchor only through the governed backend publication path. Not financial advice.
Human checkpoint
The Resistance was never meant to be a solo project. It is not a movement against artificial intelligence. It is a refusal to stop thinking simply because our machines have become extraordinarily capable.
The people leading today's AI revolution deserve admiration for what they have built. But even extraordinary products can become dangerous ideas when we fall so deeply in love with them that we confuse automation with progress and remove humanity from the decision.
Humanity should not fear machines that think, but men and women who stop doing it.
The future we want gives people better machines, better evidence and a wider search space — while keeping purpose, judgment and responsibility human. The machine can search. The human must still define the game.
We should never build a future that looks back at its own ruins and can only say:“People lived here.”Join the Resistance · Read the manifesto →