Strategic Acquisition Brief · Confidential
MaxiFi is the deterministic, consumption-smoothing optimization engine that computes the provably correct lifetime household plan — taxes, Social Security, longevity, sustainable spending — the answer a probabilistic model cannot generate by sampling tokens. Built over 30 years by BU economist Laurence Kotlikoff. The antidote Google can train directly into Gemini’s weights — computed, verifiable, reproducible answers it can stand behind.
For the AI-answer surfaces, the defining risk is no longer hypothetical. A European court has now ruled that an AI Overview is Google’s own speech — not neutral aggregation — and rejected the argument that users are responsible for fact-checking it. On the one domain where a confidently wrong answer does real financial harm — how much can I safely spend, and how do I make it last? — the defense Google currently relies on is a disclaimer. That is precisely the defense the court struck down.
A general-purpose language model is a probabilistic next-token predictor. It approximates. Separately in the U.S., Wolf River Electric is suing over a false AI Overview for $110–210M. The pattern the courts are describing — the reliability of the AI model is the live issue — maps directly onto consumer finance, where the wrong number is delivered with confidence at population scale. A disclaimer no longer contains it.
MaxiFi resolves it — the validated, deterministic engine that produces the mathematically correct lifetime plan. The antidote is not a better disclaimer; it is a computed answer Google can stand behind.
Across its consumer-finance surfaces, Google does many things well — and one thing not at all. Google Finance (rebuilt Nov 2025) issues hundreds of simultaneous searches to research markets and produce a cited response. The Gemini app answers open-ended retirement and budgeting questions in fluent prose. What no surface does is compute a sustainable, defensible lifetime household plan — the single number a household actually needs.
This is not a Gemini weakness; it is a category gap. Every large language model estimates. MaxiFi computes. A model trained on text will produce a plausible figure for “how much can I spend in retirement” and state it with confidence. Only a deterministic optimization engine, solving the household’s taxes, Social Security timing, and longevity simultaneously, returns the figure that is correct by construction. That computational household-planning primitive is the conspicuous absence in Google’s stack.
Every estimating engine answers the aspirational question — “how much will you need?” — which manufactures a target the household may never hit. MaxiFi starts from the defensible one: “what is the most I can spend with what I have?” — sustainable by construction. That is the answer a regulator, a fiduciary, and a billion users can rely on.
Google’s current financial-advice posture is pure disclaimer — “you should not rely on Gemini… as financial advice.” The May 28 ruling describes, in print, why that is no longer a shield for the AI-answer surfaces.
A preliminary injunction holding Google directly liable for false statements in AI Overviews — ruling they are “independent, new, and substantive statements,” Google’s own speech, and rejecting the argument that users are responsible for fact-checking. Up to €250,000 per violation.
Google is appealing. Cited here as a live landmark, not a final judgment — but the legal theory it validates is now in the open.
The theory is simple and it travels: if the AI answer is Google’s own speech, the “we just surface third-party content / novel technology” defense is foreclosed, and so is “the user should have checked.” Apply it to consumer finance, at 1B+ AI Mode users, and any systematic planning error becomes a population-scale liability the disclaimer cannot reach. In the U.S., Wolf River Electric’s $110–210M suit shows the same exposure already has a price tag.
The antidote is cybersecurity-style, not legal-style. A security vendor lets its customers stand behind what they ship; MaxiFi lets Google stand behind the financial answer. The response to the Munich theory is not a stronger disclaimer — it is a computed, verifiable, reproducible answer that cannot hallucinate the lifetime math, with a deterministic calculation a fiduciary and an examiner can audit. The correctness layer and the liability shield, in one engine.
MaxiFi is the financial-planning platform of Economic Security Planning, Inc., built over more than three decades by Professor Laurence Kotlikoff of Boston University. It uses consumption smoothing and dynamic programming to compute the single, mathematically optimal lifetime plan — solving simultaneously across Social Security strategy, Roth-conversion sequencing, withdrawal order, and the full current tax code.
Three planning modes are user-selectable: Safe Investing (pure deterministic — a unique solution against a guaranteed safe return); Upside Investing (deterministic floor from safe assets, Monte Carlo on risky upside); Full-Risk Investing (Monte Carlo across the full asset stack). Crucially, every Monte Carlo step runs the deterministic code — which is what lets MaxiFi properly incorporate taxes and path-specific cash flow in stochastic mode. No other Monte Carlo platform gets this right.
Estimating planners answer “how much might you need?” MaxiFi answers “what is the optimal path, and how much can I spend today without jeopardizing tomorrow?” It is not a better simulator. It is a different class of engine — the missing computational primitive.
Prof. Laurence Kotlikoff — William Fairfield Warren Professor at Boston University; Harvard Ph.D.; former Senior Economist on the President’s Council of Economic Advisers; Fellow of the American Academy of Arts & Sciences and the Econometric Society; named by The Economist among the 25 most influential economists.
Taught by Nobel Laureate Robert Merton at MIT Sloan as an “outstanding science-based lifecycle and retirement management platform” (Merton does not endorse products); featured in Bankrate’s “Best financial planning software of 2025” roundup (May 5, 2025), cited as best for near- and long-term tax planning and the decumulation phase. The economics trace to Nobel-recognized work on lifecycle consumption and optimization.
Patented algorithms refined over 30+ years, built from economic theory rather than scraped text — exactly the kind of intellectual property a large language model cannot reverse-engineer.
Larry Kotlikoff intends to keep contributing to the product, help the acquirer integrate, and continue as spokesperson. The acquirer buys the engine and keeps the economist who built and validates it — de-risking the integration.
The fastest, cleanest integration is not a runtime plug-in. It is to train Gemini on MaxiFi-generated cases — so the correct economics lives inside the model’s weights, not in a separate layer to stand up, operate, and maintain. Google alone has the two ingredients at scale: the data is public, and the compute is already in-house.
Kotlikoff has laid out the method publicly: take the Federal Reserve’s Survey of Consumer Finances — a public dataset — perturbate its observations into billions of synthetic households, run each through MaxiFi’s 30-year engine, and train on the verified input–output pairs. Gemini stops improvising on money and starts returning answers that are correct by construction.
MaxiFi is offered for acquisition. Acquiring the engine and its founder hands Google exactly the SCF→MaxiFi training program above: MaxiFi supplies the ground-truth training signal, Kotlikoff joins to integrate it, and Gemini keeps the interface, the reach, and now the math. The integration thesis and the deal are one and the same — own the engine, train the model.
Larry has published a multi-post sequence on his Substack running named LLMs — Gemini, ChatGPT, Claude, Perplexity — against MaxiFi on real household problems. Findings are dated, reproducible, and dollar-specific. The argument is never “Gemini is uniquely bad” — it is that every model estimates while MaxiFi computes.
“Gemini generated a smile that is 13 percent too low in each of Carol’s potential remaining 40 years.” Gemini computed $47,150 at 60 / $32,788 at 80 vs. MaxiFi’s correct $53,923 / $37,746.
Read the retirement-smile test →Money.com’s graders found both ChatGPT and Gemini stumbled on current data — stale by construction, because a text model retrieves rather than computes. The category property, not one model’s flaw.
The pattern, not the outlier →“AI’s best hope of providing accurate economics-based planning is by… using MaxiFi as a back end to produce precisely correct, not clearly pretend results.” The structural argument for computing over estimating.
Read the structural argument →At least one benchmark has Gemini outscoring competing models on financial planning. We make the honest argument: every LLM estimates; MaxiFi computes. The 13%-too-low miss is the category, not the model.
Why the category matters →Larry’s Substack has 137,000+ subscribers as of May 2026 and is growing. Acquiring MaxiFi acquires the megaphone these pieces ship from — pointed, with credibility no one in the category can match. Larry’s own architectural recommendation is the integration path on the table: use MaxiFi and the Federal Reserve’s Survey of Consumer Finances to train Gemini on the correct deterministic math — so the math becomes native to the model rather than an external dependency.
The Munich ruling and the Wolf River suit describe Google’s AI-answer exposure at billion-user scale, and the current defense — a disclaimer — is the one the court rejected. MaxiFi is the antidote: computed, verifiable, reproducible answers Google can stand behind on the one domain where a wrong answer costs.
The training data is public (the Survey of Consumer Finances) and the compute is already in-house. Google is the one buyer that can absorb MaxiFi’s verified cases into Gemini’s weights at scale — no separate runtime layer, no new compute line item. Ease and speed are the point.
Google Finance researches markets; Gemini estimates plans; neither computes a sustainable lifetime answer. MaxiFi is that missing primitive — the computational household-planning capability absent from the stack today, dropped into the surface with 1B+ users.
The acquisition brings the architect with it: Kotlikoff stays on to integrate the engine, validate the training program, and continue as spokesperson. The acquirer owns the engine and keeps the economist who built it — so the SCF→MaxiFi training path is staffed by the person who designed it.
There is one MaxiFi. If it lands at OpenAI, Anthropic, Meta, or a large fintech, Google’s claim to the correct-by-construction financial answer weakens permanently — the most defensible single piece of AI-FS infrastructure owned by a rival. The denial case sits alongside the offensive one.
A 30-minute briefing with a live demonstration — MaxiFi solves “Carol’s” lifetime plan while Gemini is asked to match it. The gap is the entire thesis.
MaxiFi is being offered for acquisition through a focused strategic process. For Google the integration is short, the data is public, the compute is in-house, and the strategic payoff — a named-threat antidote, the missing computational primitive, and competitive denial — is immediate. Owning the engine and its founder is the train-the-model path.