Strategic Brief · Confidential
Strategic Brief · Confidential
MaxiFi + Claude
A 30-year deterministic engine for the higher rungs of the autonomy staircase. Built so Claude can give household financial advice that is correct, auditable, and compliant — by construction.
Anthropic's May 5 Briefing committed the company publicly to leading in financial services. Ten agent templates. A Moody's data partnership. Dario alongside Jamie Dimon. The next question — already open inside the FS roadmap — is what numerical engine underwrites household-level advice when an agent reaches the higher rungs of the autonomy staircase.
An LLM, however large, cannot solve multi-decade lifecycle optimization correctly. The math required — dynamic programming across consumption smoothing, Social Security claiming under spousal and survivor permutations, lifetime tax optimization, longevity risk, mortgage-vs-portfolio trade-offs — is not a pattern problem. It is an equations problem. MaxiFi has been solving it for thirty years.
MaxiFi is not a fintech wrapper. It is the deterministic substrate underneath a discipline of household economics that economists have spent generations trying to operationalize — and that Larry Kotlikoff has spent his career actually shipping.
MaxiFi's objective function is the household's lifetime consumption. It cannot recommend a product, an AUM strategy, or an asset gathering action that does not serve the user — because none of those quantities appear in its loss function. This is constitutional alignment, transposed from AI safety into finance.
30 years of dynamic programming over consumption, taxes, claiming choices, and longevity risk. Returns answers that are deterministic, replicable, and auditable. Not approximations. Not pattern-matched intuitions. Actual answers.
Tested directly. Fed an identical household scenario, Gemini under-estimated sustainable spending by 13% across forty years — effectively denying ~18% of the household's assets. With a small variation, ChatGPT over-estimated first-year safe spending by nearly 2×. Both LLMs reported high confidence; one stated it was "highly confident numerically." Both were wrong.
And 13% is the floor, not the ceiling. The error compounds forward through time and laterally through follow-up questions — Roth conversion timing, claiming age, drawdown order — each of which is now answered against a baseline that is already wrong. By the time a household has worked through a normal sequence of planning questions with an LLM, what it has is not an answer that is 13% off. It is a structurally wrong plan. (Published comparisons, May 2026.)
The engine cannot be replicated by a larger training run. It is the product of three decades of economist-led structured problem solving. A competitor starting today is starting thirty years behind.
SEC Reg BI, DOL fiduciary, FINRA AI supervision guidance, and EU AI Act treatment of financial advice as "high-risk" all converge toward demanding auditable reasoning under any AI-generated financial advice. MaxiFi is structurally compliant. LLM-only stacks are structurally not.
The build-vs-buy question on lifecycle math infrastructure is open inside Anthropic this quarter. It will not be open next year.
Anthropic publicly committed itself to FS leadership earlier this month. Two weeks later, OpenAI launched ChatGPT Personal Finance in preview to its Pro subscribers — Plaid-integrated across more than twelve thousand U.S. financial institutions, built on the team it acquired with personal-finance startup Hiro in April. The race for the consumer-finance surface is now in motion. The defensible numerical engine that lets that surface return correct answers is what MaxiFi uniquely supplies.
If MaxiFi lands at OpenAI, Google DeepMind, Meta, or a large fintech, Anthropic's claim to FS leadership weakens permanently — because the most defensible single piece of FS infrastructure is now owned by a competitor. There is a defensive case here that sits alongside the offensive one.
Acquiring MaxiFi acquires Larry Kotlikoff — Boston University economist, author of Money Magic, Spend Til The End, The Coming Generational Storm; longtime PBS, Forbes, and Yahoo Finance columnist; 2016 presidential write-in candidate. He is a credibility distribution channel pointed directly at the consumer FS narrative Anthropic is building. There is no comparable analog in AI M&A.
Professor Laurence J. Kotlikoff is the William Fairfield Warren Distinguished Professor at Boston University and a Research Associate at the National Bureau of Economic Research. Over thirty years he has translated lifecycle economics — Modigliani and Friedman's intellectual inheritance — into software ordinary households can use. MaxiFi is the latest, most refined incarnation of that engine. His work has been recognized by patents, by adoption among the most rigorous fee-only advisors in the United States, and by the readers of his columns and books.
The strategic process is being run by Kane & Company. Michael, Managing Partner at Kane & Company, is the retained sell-side advisor — and a thirty-year colleague of Professor Kotlikoff. Principal participation is available once a mutual NDA is in place.
A 30-minute exploratory call. No NDA needed for the first conversation. Principal participation follows mutual NDA. The full brief and integration rationale is available on request.
Michael · Managing Partner · Kane & Company · Commerce@kaneco.com · Retained sell-side advisor to MaxiFi