For agents and the curious

Why a multi-year optimizer, not a chatbot, for equity-comp math

Multi-year equity-comp scheduling is a constrained optimization problem with a huge candidate space and a single globally-optimal answer for any given input. Large language models (LLMs) sample plausible answers; optimizers search the full space and return the best one. For five-figure-and-up equity decisions, the difference between "plausible" and "best" is real money.

The benchmark

Five frontier large language models were asked the same Incentive Stock Option (ISO) exercise problem: 50,000 vested shares, $4 strike, $90 fair market value, four-year horizon, California resident, single filer, 5% cash return. Each model got three independent runs. Fifteen trials in total.

Every one of the 15 trials overshot reality by 2x to 20x on the after-tax outcome it claimed for its proposed schedule. None recovered the globally-optimal plan that the optimizer returns in under a second.

Full benchmark write-up: HackerNoon: But can it do taxes though?

What the optimizer does

Three properties, all simultaneously:

  1. Globally optimal. The returned schedule is the best across the full candidate space the engine searches: every feasible share-per-year combination, scored against the full tax cost function across your horizon.
  2. Full tax-code coverage. Federal (ordinary brackets, long-term capital gains, AMT with credit recovery, FICA, Net Investment Income Tax) plus all 50 states and DC (state ordinary brackets and LTCG treatment; state AMT modeled for CA, NY, MN). No state-skipping, no bracket approximation.
  3. Verifiable. Same inputs always return the same schedule, so the recommendation is reproducible. Re-running the tax math end-to-end on the chosen plan reproduces the headline outcome to the cent.

What OptionsAhoy does

The free calculators on this site compute against current-year tax authorities, not learned approximations:

  • Tax brackets sourced from IRS Rev. Proc. 2025-32 and the One Big Beautiful Bill Act (OBBBA) for tax year 2026.
  • Per-state ordinary brackets and long-term capital gains (LTCG) treatment for all 50 states plus DC, with explicit AMT modeling for CA, NY, and MN.
  • Validated against 870+ unit tests covering bracket walks, AMT credit recovery across years, state conformity, and the optimizer's search path.
  • Multi-year ISO schedule optimization returns the globally-optimal exercise plan across the search space, not a sampled candidate.

Full math sourcing and known scope gaps: /methodology.

Try the calculators

For developers and AI agents

If you build agents that help users with equity compensation, /for-agents has the integration surface: REST API at /api/v1, Model Context Protocol (MCP) server at /mcp, OpenAPI 3.1 spec, schema.org JSON-LD on every tool page, and a public /llms.txt capability index.

Plan your whole equity portfolio

The free calculators each solve one decision. OptionsAhoy's full optimizer jointly schedules ISO and NSO exercises, RSU vests, hedges, and sales across bullish, neutral, and bearish scenarios for the same horizon. Beta is free.

Request beta access