Equity to Cash at the Risk You Set

Optimization. Bracket-aware greedy fill across every (sale-year, sale-month, lot) cell from now through the target date. Each candidate block is priced by its full incremental tax (federal long-term and short-term capital gains, state, and NIIT, per the running year-state) and then grown forward at your cash-interest rate to the deadline. At each step it picks the block with the highest conversion benefit: after-tax-and-interest cash now, minus what those shares would be worth held to the deadline. That funds your goal from the holdings least worth keeping, such as a position projected to decline, while retaining the ones expected to grow. It refines 100 → 10 → 1 share blocks until the target is hit; on single-stack inputs the loss vs the true wealth-maximizing optimum stays within 1–2%. The search is deterministic: the same inputs on the same date return the same answer, so the recommendation is verifiable.

Tax-code coverage. Relevant federal code (ordinary brackets, long-term capital gains, AMT with credit recovery, FICA, Net Investment Income Tax) plus each state's own return for all 50 states and DC (starting income, deductions, exemptions, surtaxes, LTCG treatment). State AMT modeled for CA, CO, CT, MN. Brackets reflect IRS Rev. Proc. 2025-32 and the One Big Beautiful Bill Act (OBBBA) provisions for tax year 2026. Validated against 870+ unit tests covering bracket math, AMT credit recovery, state conformity, and the optimizer's search path.

Why use an optimizer rather than asking an LLM to do the math: a benchmark of five frontier large language models on the same multi-year ISO problem found that all 15 trials overstated the after-tax result of their own proposed schedule by 2x to 20x (HackerNoon write-up).