Field 01 / Markets · Research Appendix 01

The Economics of Sufficient Intelligence

The machine searches. The human defines the game. The question is not who uses the biggest model, but whether extra intelligence produces enough decision value to justify its cost.

Falsifiable research question

Can human-directed AI produce economically competitive investment outcomes without maximizing computational expenditure?

We compare the cost structures of five decision architectures. Better model scores do not always imply greater net economic value. Higher spending may nevertheless be justified when its marginal benefit exceeds its marginal cost.

Four histories, one explicit proxy

DTRM and ChatGPT Fund use governed portfolios of eleven shares, with decisions around every 60 days. The published experiment also follows a S&P 500 benchmark and AIEQ, a dividend-adjusted AI-managed ETF.

Qwen Local is not an investment track record. It assumes the very same cumulative return as ChatGPT Fund solely to compare the economics of owning a two-GPU local workstation with contracting a model subscription.

These choices do not establish that human judgment causes higher returns, or that a portfolio is tradable at institutional scale.

Evidence status

Published histories
4
Proxy scenarios
1
Through
2026-11-13
Snapshot
Checked-in

The expense model is declared and scenario-based; there is no audited historical cost ledger.

Reference capital · USD

ChatGPT Fund reference: Pro 200 · $200/month. Qwen's hypothetical trading path follows ChatGPT Fund. No Plus scenario.

Priced annual items · not total execution costs

This compares published ETF expense ratios with stated operating budgets and modeled Lightyear order commissions. Not total trading costs.Real execution prices, regulatory sale fees, and order-size acceptance remain unverified.

DTRM Fund$630 / year0.06297% of capital

Technology: $529 · Broker: $100.71 · Other execution costs: not quantified

ChatGPT Fund$2,470 / year0.24702% of capital

Technology: $2,401 · Broker: $69.16 · Other execution costs: not quantified

Qwen Local (proxy)$2,569 / year0.25692% of capital

Technology: $2,500 · Broker: $69.16 · Other execution costs: not quantified

S&P 500 (SPY Total Return)$945 / year0.09450% of capital

Embedded ETF expense: $945 / year

AI Managed ETF — AIEQ$7,500 / year0.75000% of capital

Embedded ETF expense: $7,500 / year

Cohort target-weight deltas imply about 8.43× capital traded annually for DTRM versus 3.79× for ChatGPT. These are gross buys + sells from frozen portfolio targets, not observed order fills or actual post-price-drift turnover.

The initial purchase of eleven stocks is separate from the recurring model. Assuming each order is accepted in one piece, its published broker commissions would be DTRM $11 · ChatGPT $11 · Qwen $11. This does not include actual fill costs, regulatory charges, FX or applicable tax.

ETF expenses are reflected in fund NAV and must not be deducted a second time. Nominal broker fees exclude FX, dividend taxes, applicable regulatory sale fees, ADR charges, order splits, delays, spread and nonlinear market impact.

Lightyear publishes no universal maximum equity-order notional that we could verify. At $1m, $100m or $1bn, displayed broker commissions are conditional on single-order acceptance. These are not total executable trading-cost estimates. Actual execution capacity and cost require broker confirmation and observed market data.

Trading commissions · different portfolio rotation

All figures below cover the same 507-day historical observation window beginning 17 June 2025. They include the initial eleven purchases and both sides of eight portfolio rebalance events. Qwen uses ChatGPT's trading path by assumption, not observed Qwen decisions.

DTRM Fund

Frozen target-weight changes

Initial buy orders
11
Rebalance buy orders
72
Rebalance sell orders
68
Gross buy notional / capital
5.85×
Gross sell notional / capital
5.85×
Initial buy commissions
$11.00
Rebalance buy commissions
$72.00
Rebalance sell commissions
$67.80
507-day total nominal commissions
$150.80

ChatGPT Fund

Frozen target-weight changes

Initial buy orders
11
Rebalance buy orders
45
Rebalance sell orders
51
Gross buy notional / capital
2.63×
Gross sell notional / capital
2.63×
Initial buy commissions
$11.00
Rebalance buy commissions
$45.00
Rebalance sell commissions
$51.00
507-day total nominal commissions
$107.00

Qwen Local (proxy)

Proxy: ChatGPT trades

Initial buy orders
11
Rebalance buy orders
45
Rebalance sell orders
51
Gross buy notional / capital
2.63×
Gross sell notional / capital
2.63×
Initial buy commissions
$11.00
Rebalance buy commissions
$45.00
Rebalance sell commissions
$51.00
507-day total nominal commissions
$107.00

Commission model: 0.1% per single filled US-listed share order, with $0.10 minimum and $1 maximum. Costs apply symmetrically to modeled buys and sells. Figures rely on target-weight changes, not actual execution records, and do not include regulatory sale charges, ADR or FX costs. They are conditional on one accepted order per position change even at $100m and $1bn; no maximum permitted Lightyear equity order size is confirmed. Changes between reference and fill prices can help or hurt a trade. No price-deviation distribution or arbitrary execution penalty is assumed.

Historical return · mechanical capital illustration

2025-06-17 → 2026-11-13

Four published market histories, one Qwen proxy assumption. Values below apply each reported cumulative return to the selected capital without accounting for fund capacity, real-world execution, FX, dividend tax or personal fees. These are not investable NAV estimates.

DTRM Fund

Published
Cumulative return
+48.76%
Illustrative ending value
$1,487,584

ChatGPT Fund

Published
Cumulative return
+30.29%
Illustrative ending value
$1,302,868

Qwen Local (proxy)

Proxy · hypothetical
Cumulative return
+30.29%
Illustrative ending value
$1,302,868

S&P 500 (SPY Total Return)

Published
Cumulative return
+28.34%
Illustrative ending value
$1,283,444

AI Managed ETF — AIEQ

Published
Cumulative return
+18.45%
Illustrative ending value
$1,184,456
Read the return basis of each strategy
DTRM Fund
Dividend-adjusted equity closes; not after individual tax or actual execution.
ChatGPT Fund
Dividend-adjusted equity closes; not after individual tax or actual execution.
Qwen Local (proxy)
Proxy assumption: identical to ChatGPT; not a Qwen track record.
S&P 500 (SPY Total Return)
SPY dividend-adjusted total-return ETF proxy: reinvested dividends and fund expenses included; not the official gross S&P index.
AI Managed ETF — AIEQ
AIEQ dividend-adjusted ETF series; ongoing fund expenses reflected in prices.

Return basis: DTRM Fund, ChatGPT Fund and AIEQ use dividend-adjusted closes; S&P 500 uses SPY dividend-adjusted total return (dividends reinvested, fund fees reflected). Performance and risk ratios exclude investor-specific taxes and trading fees.

Expense assumptions and units

DTRM: $528/year of operator-declared FMP ($228), GitHub ($48), Netlify ($108) and allocated ChatGPT ($144). Its fixed operating estimate is $529/year including $1 of occasional home electricity. Lightyear commissions are shown separately for each modeled target-weight order. Actual fill prices, slippage, market impact, regulatory fees and maximum accepted equity order sizes remain unverified. The former $135 flat estimate is retired.

ChatGPT: Pro 200 is a $2,400/year subscription scenario plus $1 annual electricity. Trading follows the same order-price model as DTRM, but ChatGPT has lower observed target-allocation turnover. This remains a scenario, not an audit of historically incurred Pro subscriptions.

Qwen Local: hypothetical two-GPU NVIDIA workstation at $9,000 upfront, depreciated over four years ($2,250/year), plus $228 financial data and $22 electricity. Trading uses the ChatGPT Fund target-allocation path as a hypothetical proxy. No Qwen portfolio performance has been observed.

ETFs: 0.0945% annual fund expenses for the SPY S&P 500 ETF proxy; 0.75% for AIEQ. Their operating fees are not equivalent to the cost of running an investor's decision model and are generally reflected in ETF NAV.

Scientific boundaries

H1 · Cost efficiency. Compare recurring budgets; this is sensitive to subscriptions and hardware configuration.

H2 · Economic outcomes. Dividend withholding, actual trading and instrument-level fees require reconciliation before net-performance claims.

H3 · Decision scalability. Test how fixed costs dilute at $1m, $100m and $1bn; not a test of liquidity or executable capacity.

H4 · Sufficient intelligence. Requires future controlled model-quality and resource-consumption tests. Not demonstrated here.

We exclude personal investor time consistently, but not all institutional obligations. Expenses, market impact, FX and taxes are not harmonized.

Methods, evidence and further reading

Cumulative returns are derived dynamically from the same governed four-series market snapshot used by InvestGame. The original performance and MM1 are untouched. Source dates: 2025-06-17 to 2026-11-13. Historical capital values are arithmetic illustrations, not tradable NAVs, not forecasts and not adjusted for actual execution.

Displayed DTRM/ChatGPT/Qwen annual charges = operating budget + annualized modeled Lightyear order commissions. Fees per modeled US share order are capped at $1 with a $0.10 minimum, conditional on the broker accepting a single order per target-weight change. No hypothetical basis-point fee is inserted. Actual trading cost and order feasibility are unknown, and therefore not included in the annual subtotal. For ETFs, the displayed annual fund expense is chosen capital times the expense ratio; cost percent is displayed annual amount divided by reference capital. The S&P 500 series is a SPY ETF dividend-adjusted total-return series (with ETF fees reflected), not the official gross S&P 500 index. AIEQ uses adjusted closes. The SPY fund expense shown above is already reflected in the adjusted-close path and is not deducted again.

Contract research-appendix-01-v0.2. Dividend-event reconstruction, tax, FX, commission reconciliation, investable ETF tracking and compute-metering, observed fill/market-impact measurement and real broker order-size validation remain explicitly pending.

Evidence: public market snapshot. External conventions: Lightyear, SPY issuer, AIEQ, OpenAI pricing.

The most powerful intelligence is not always the most economically valuable. Research only; not financial advice.