SI KILLED THE GDP STAR (IN THE LONG RUN) — THE LAST HBS CASE BEFORE THE BELL — NueraPump presents SI Killed The GDP Star, a seven act House courtroom musical and interdisciplinary research proposal examining whether synthetic intelligence can expand productive capacity while weakening the income circulation required to sustain realized output. A fictional Harvard classroom becomes a Yuan drama tribunal: Guan Hanqing 关汉卿 presents June's household bills, Gold Teeth 金牙大状 defends Clark's growth scenario, and Principal Maverick 卞院长 requires both advocates to expose assumptions, trace consequences, and accept contrary evidence. The classroom setting claims neither Harvard endorsement nor actual legal proceedings. Its central distinction separates potential productivity, realized production, household purchasing power, and welfare. Competitive adoption may be rational for each enterprise without guaranteeing aggregate stability. Lower task costs can change prices, employment, wages, profits, and market shares; their combined consequences depend on recipients, spending propensities, timing, domestic production, debt, and policy. The IPO dispute examines valuation defense through principal agent analysis. Reported investor discussions moving from approximately two trillion dollars toward one and a half trillion indicate pricing pressure, not a finalized offering valuation. A planned listing amid financing uncertainty invites investigation of liquidity needs, investor exit, dilution, and contractual obligations; it does not independently establish deception. Macroeconomic expansion is not automatically corporate addressable revenue, and addressable revenue is not distributable cash. Valuation must connect paying customers, sustainable prices, retention, operating margins, reinvestment, working capital, financing claims, and discounted future cash flows. Rapid revenue growth need not produce comparable shareholder returns when competition compresses multiples or financing dilutes ownership. Loss measurement distinguishes operating losses, financing instrument revaluations, capital expenditures, and actual cash consumption. Large noncash charges cannot be described as equivalent cash outflows, but removing them does not eliminate operating or funding constraints. Alphabet's quarterly negative free cash flow enters as evidence of investment exceeding contemporaneous operating cash generation, while positive trailing annual cash flow remains visible. CoreWeave's repricing illustrates sector financing concerns without proving identical exposure across businesses. The relevant question is whether committed payments can be met if issuance or refinancing is delayed. Five analytical lenses structure the confrontation: principal agent theory examines incentives and risk allocation; value chain analysis traces domestic value creation and customer payments; negotiation examines bargaining alternatives and institutional delay; double loop learning challenges the governing objective; balanced scorecard reasoning separates output, household conditions, debt, and public finances. None substitutes for causal estimation. The macroeconomic teaching model assumes fixed prices, spare capacity, no government, no foreign trade, and consumption C=[c_Ls+c_K(1-s)]Y, yielding Y=A/{1-[c_Ls+c_K(1-s)]}. With c_L=0.9, c_K=0.4, autonomous investment A=30, and labor share s=0.6, equilibrium output equals 100; reducing s to 0.4 gives 75. Raising A to 40 restores 100, while 50 yields 125. These are illustrative units, not American forecasts. Saving still equals investment; no accounting dollar disappears. The example demonstrates a conditional contractionary possibility while leaving adoption, distribution, prices, investment, and policy to be determined in a fuller model. National accounting imposes additional boundaries. Y=C+I+G+ΔX-M records production through final expenditure, not a guarantee of full employment. Domestic outsourcing can move value added from clients to cloud suppliers without reducing national value added. Imported production must be excluded without repeatedly subtracting foreign supply chain components. Gross operating surplus already includes capital consumption; depreciation cannot be added twice. Investment shares differ from incremental growth contributions, global corporate spending differs from domestic capital formation, and stock appreciation differs from newly produced output. Neither wage losses nor transfers generate a universal sevenfold multiplier. Three outcomes require separate tests: nominal compression, real output below a credible counterfactual, and sustained absolute real contraction. Attributing the third to automation additionally requires identifying its contribution and reinforcing feedbacks. The reported 44.4 trillion scenario uses constant 2025 prices and is neither an IMF forecast nor an assigned probability. Its omitted demand feedback warrants investigation, not automatic rejection or substitution with an unestimated 25 trillion outcome. Dynamic evaluation follows households, firms, government, banks, and overseas accounts quarterly, separating consumption groups, investment types, imports, taxes, transfers, borrowing, principal repayment, and defaults. Fast replacement and slow purchasing power reconnection define a stress configuration, not a sufficient collapse condition. Delayed assistance may permit business exits, skill loss, and debt distress; their persistence and repair costs require measurement. New tasks, lower prices, broader ownership, investment, and public services must enter as quantified responses rather than promised rescue. Human Settlement remains a lawful distribution proposal requiring resources, acceptance, clearing, accountability, and accessible dispute resolution, with basic protection independent of technical scores. Measurement uses Q=dS/dt as shorthand for verified change in a declared state mapping per specified time. Operationally, Q=[V(S_verified(t1))-V(S_verified(t0))]/(t1-t0), with goals and constraints fixed before testing. Change alone is not improvement. Corporate efficiency cannot cancel household harm through an arbitrary average; elapsed time, human effort, machine resources, uncertainty, and reversals remain distinct. The NeuraPump educational method translates these relationships into memorable characters, images, rhythms, and synchronized actions. House repetition carries retrieval cues, while the fourth act's silence exposes wages disappearing before rent obligations. Apate's earlier distinction between activity and protection becomes the courtroom distinction between impressive forecasts and settled benefits. Fantasy supplies expressive surfaces while real mechanisms constrain consequences. Early access and repeated exposure aim to develop analytical habits, but neither thirty viewings nor thirty thousand case maps guarantees executive competence. Delayed recall, independent explanation, novel cases, and correction under changed constraints test transfer. LyricManga treats each illustrated page as a case environment whose spatial arrangement, colors, expressions, and action sequences expose objectives, information asymmetries, and causal dependencies. Lyrics, synopsis, and visual instructions guide model completion of scene details without authorizing invented evidence. A shared timeline coordinates meaning across voice, image, gesture, and emotional reversal; fifty millisecond scheduling describes an engineering resolution, not proof of controlling every muscle or measuring learning at that frequency. Fishbone diagrams organize candidate causes, while counterexamples prevent familiar pictures from becoming unquestioned answers. The intended inheritance is practiced judgment across unfamiliar real situations, not institutional prestige. The final question remains open: who can purchase the promised future, who finances the transition, and what evidence would make either advocate revise the verdict?