How the math works

We grow each account (401(k), Roth, savings, brokerage) separately using the future value of an annuity-due formula, accounting for both your existing balance and ongoing contributions.

r = pre-retirement return, n = years to retirement. Lump sums (home sale, inheritance) are grown to retirement age. Household plans grow each spouse independently then combine at your retirement date. In retirement, a household plan runs to whichever spouse lives longer — set each spouse's life expectancy separately — and switches to survivor economics at the first death: spending drops to your survivor goal (default 75% of your monthly goal), the larger of the two Social Security checks is kept (the smaller stops), tax filing becomes single, and half of the deceased spouse's pension continues while the survivor's own pension continues in full. When the two life expectancies differ, this materially changes projected legacy and survivor planning.

Your benefit is calculated from your Full Retirement Age (FRA) benefit and adjusted for when you claim. Claiming before FRA permanently reduces it; delaying past FRA increases it up to age 70.

SSA factors: ~0.70 at 62, 1.00 at FRA (67 for most), 1.24 at 70. This tool assumes a Full Retirement Age of 67, which applies to anyone born in 1960 or later. If you were born before 1960, your FRA is slightly lower (66 for born 1943–1954, graduating to 67 by 1960) — enter your actual FRA benefit to keep the math accurate regardless. Spousal benefit is the higher of their own record or 50% of your FRA benefit. Benefits are inflation-adjusted to retirement-year dollars. Stress-testing for legislative shortfall: the SSA Trustees project the trust fund hitting depletion in 2033, after which scheduled benefits would be reduced by ~20-23% absent congressional action. To stress-test your plan against this, set the "SS Benefit Confidence" slider to 80% (or another haircut you find plausible) — the engine will scale all SS benefits accordingly.

Your Monthly After-Tax Income headline is the spend level your projection actually sustains through your full retirement horizon. We compute it by bisecting the deterministic projection: probe a spend amount, run the year-by-year simulation, check whether the portfolio depletes; raise the probe if it survives, lower it if it depletes. The result is the maximum monthly spend you can fund without running out — a readout of the projection itself, not a separate model.

The Monte Carlo success rate (Section 6) tells you the probability of that draw working across 1,000 different market scenarios — use the deterministic headline for capacity, the MC rate for confidence. The 4% rule was calibrated to a 30-year horizon with ~95% historical survival; it was not designed for someone retiring at 55 with a 35+ year horizon, and it doesn't account for guaranteed income streams (Social Security, pensions) that change the math materially. The bisection answer respects your actual horizon and your actual income streams.

We model post-retirement taxes using 2026 federal brackets (inflation-adjusted forward to each future year), state-specific tax codes for all 50 states plus DC (see Section 19), and the IRS Publication 915 provisional income formula for Social Security taxability. Tax is computed on your full income each year — pension, taxable SS, ordinary 401(k) draws, and long-term capital gains on brokerage draws. Brokerage gains count in your AGI when determining how much of your Social Security is taxable, which materially affects retirees with significant taxable accounts. Both the deterministic projection and Monte Carlo simulations use tax-aware withdrawal ordering and stratified LTCG. The provisional-income thresholds themselves ($25,000/$34,000 single, $32,000/$44,000 married) are held FIXED in future years, matching the statute — Congress has never inflation-indexed them, which is why a growing share of Social Security becomes taxable over a long retirement — while tax brackets, the standard deduction, and capital-gains breakpoints do index forward. The engine also applies the 3.8% Net Investment Income Tax on realized gains and cash interest above the statutory $200,000/$250,000 MAGI thresholds — which the law also leaves unindexed — the additional standard deduction from age 65 (per person, indexed, and correctly reduced to the single amount in survivor years), and ordinary income tax on high-yield savings and bridge-reserve interest in the year it accrues (interest the model previously grew tax-free). For tax years 2025–2028 the engine also applies the temporary OBBBA senior deduction: $6,000 per person aged 65+, reduced by 6% per person of income above $75,000 single / $150,000 married — amounts the statute fixes, so they are never inflation-indexed here. It is a below-the-line deduction: it lowers taxable income but never changes Social Security taxability, NIIT, Medicare IRMAA, or ACA subsidy income. It expires after 2028 and the projection drops it then automatically.

401(k) draws are ordinary income. Brokerage draws are taxed at stratified long-term capital gains rates (0% / 15% / 20%) — the rate depends on where the LTCG sits relative to your other income that year, not a flat 15%. Roth withdrawals are tax-free. Savings (HYSA) withdrawals are also tax-free — principal is treated as already-taxed cash; interest is assumed taxed annually as it accrues, consistent with real-world HYSA behavior. Tax brackets and the standard deduction inflate forward each year (when "Inflation-adjust brackets" is enabled in your inputs), matching the inflation applied to your spending and balances — so your real tax burden tracks consistently across the projection. The tax estimate shown each year combines baseline tax (owed on guaranteed income alone) plus marginal tax (additional tax caused by any portfolio draw or RMD). For households with strong guaranteed income — pensions, multiple SS streams — baseline tax can be $20-50K/yr even when no portfolio draw is needed. Many free retirement calculators skip this entirely, inflating end balances by $30-80K/yr. IRMAA (Medicare Part B + D Income-Related Monthly Adjustment Amount): when MAGI exceeds bracket thresholds (2026: $109K single / $218K MFJ for tier 1, scaling up to $500K / $750K for tier 5), Medicare premiums add a per-person surcharge ranging from ~$1,148 to ~$6,936/yr. We model these surcharges in healthcare cost for any year a household member is 65+ and the household's MAGI crosses a tier threshold; thresholds inflate at general CPI to track real-world bracket creep. Documented simplifications: we use current-year MAGI rather than the IRS's 2-year lookback, and the MAGI estimate assumes draws are ordinary-taxable (slightly overestimates for Roth/LTCG-heavy plans). Brokerage cost basis: long-term capital gains tax is computed on the realized GAIN portion of a brokerage withdrawal — not the full withdrawal amount. The engine tracks your brokerage cost basis (original principal + ongoing contributions + any lump sums routed to brokerage) and computes realized gain as withdrawal × (1 − basis/balance) when you draw. Pre-May 2026 the engine simplified by treating full draws as gain, which over-stated LTCG tax for plans with substantial taxable brokerage; the May 16 update corrected this and shifted projections up for brokerage-heavy plans by $40K–$550K at life expectancy. State tax modeling is state-level only — county, city, and local income taxes (e.g., NYC, Indiana counties, Ohio cities, Kentucky counties, MD/MO/AL/IA/MI municipalities) are not currently modeled and may add 0.5%–3.5% in affected jurisdictions. If you live in a high-local-tax area, consider bumping the "State tax %" input upward to approximate your combined burden.

The IRS requires minimum annual withdrawals from pre-tax accounts. Under SECURE 2.0, your start age depends on your birth year: 73 if born 1951–1959, 75 if born 1960 or later. The amount is your projected balance divided by an IRS life expectancy factor that decreases each year.

We use the IRS Uniform Lifetime Table (2022+) and SECURE 2.0 Act start-age rules. RMDs stack on top of SS and portfolio draws, potentially pushing you into higher brackets. Roth accounts have no RMDs. When your RMD exceeds your spending need (common for households with strong guaranteed income), we compute the marginal income tax on the overage using your year-specific bracket position — including any capital gains bracket cliff your other income causes — pay the tax, and reinvest the after-tax remainder into your brokerage account, preserving wealth that would otherwise be unnecessarily drawn down. Household plans run each spouse's share of the pooled pre-tax balance on that spouse's own clock: your share follows your birth-year start age and your attained age's divisor, your spouse's share follows theirs — so an older spouse's required withdrawals begin when THEY reach their start age, which can be years before yours. The share split is fixed at retirement in proportion to each spouse's pre-tax balances (draws and conversions deplete both shares proportionally — a simplification; real households may drain one spouse's accounts first). Two disclosed edges: a spouse already past their start age before the projection begins has their pre-retirement RMDs unmodeled (forced draws start at the first retirement year), and the survivor analysis keeps the surviving person's own schedule without modeling inherited-IRA elections.

We run 1,000 simulations of your retirement, each with randomized annual returns drawn from a log-normal distribution. The success rate is the percentage where your portfolio survives to life expectancy. Results are hypothetical and do not reflect actual investment results — they depend on your return and volatility assumptions.

μ = ln(1 + mean return) − σ²/2, σ = volatility. Random draws use the Box-Muller transform. Inflation is drawn too: each simulated future carries its own inflation path — persistent year to year (a spike is followed by elevated years, not an instant return to normal), mean-reverting to your stated rate, and mildly negatively correlated with the same year's market return, so the joint bad year — poor returns while prices rise — genuinely occurs in the sample. Your stated rate is the typical year (the median of the distribution); occasional spikes pull the long-run average slightly above it, exactly as lived history does. The process constants (persistence 0.68, shock 2.13%/yr, return correlation −0.30, floored at −1% deflation) are fitted by ordinary least squares from the same 1948+ CPI-U series the Historical Back-Test replays — public data, standard method — and an independent published fit of the same data lands on the same numbers. The year-by-year projection table deliberately stays at your stated rate: its job is one legible future you can check by hand; the Monte Carlo's job is the thousand around it. The High Inflation Decade stress scenario shifts the whole process to average 7% with its variance intact — the way the 1970s actually behaved — and the Historical Back-Test replays actual year-by-year CPI-U joined to the same years' real market returns. 85%+ is a common benchmark for 30-year retirements — for longer horizons (40+ years), many planners target 95%+. Default volatility (σ) is 12% — consistent with a balanced 60/40 portfolio. You can adjust this in the Inputs tab under Return Assumptions. Higher volatility widens the range of outcomes and lowers your success rate; lower volatility narrows it. Savings (HYSA) balances are NOT subject to market volatility in the simulations — they grow at your entered savings rate deterministically each year, since cash doesn't experience equity drawdowns. This matters for plans with significant cash positions: many calculators (including ours, prior to April 2026) accidentally subjected cash to market swings, distorting both upside and downside scenarios. Limitation: returns are modeled as independent year-to-year using your fixed mean and volatility assumptions. If future market conditions differ significantly from your inputs, outcomes will differ. Use the stress test and try lower return assumptions to understand your plan's sensitivity.

When you run a deep AI analysis, the success ring updates from a raw Monte Carlo percentage to a Plan Score — a holistic assessment that weighs your simulation results alongside factors the Monte Carlo doesn't model: sequence-of-returns vulnerability during your bridge period, Social Security timing risk, account concentration, and the sustainability of your spending assumptions.

A plan with 95% Monte Carlo success but significant sequence-of-returns exposure during an 8-year bridge period may score 85 — not because the math changed, but because the AI identified risks the simulation masked. The score uses the same health labels as the AI analysis: Excellent, Strong, Solid, At Risk, Critical. If you haven't run a deep analysis, the ring shows the raw Monte Carlo success rate.

Each year — both in the deterministic projection and in every Monte Carlo simulation — withdrawals follow a fixed, tax-aware order: any earmarked bridge reserve first (during the gap before SS starts), then savings (zero tax — already-taxed cash), then brokerage (long-term capital gains rates, stratified across 0/15/20% brackets), then pre-tax 401(k) as ordinary income, then HSA at age 65 or later (treated as ordinary income for non-medical use, like a traditional IRA), then Roth last at zero tax. Pre-65 HSA only funds qualified medical expenses (tax-free). As accounts deplete the tax burden shifts naturally — we model this rather than holding the Year 1 account mix fixed for 40 years. RMDs from your applicable start age (73 or 75 — see Section 5) are enforced as a 401(k) floor on top of voluntary draws. This conventional order is not always the lowest-tax one: drawing some pre-tax money in the low-income early years can pay less tax over a lifetime, so the Withdrawal Order card also runs your plan drawing Roth first and drawing evenly from every account, and shows the engine’s result for each.

When your RMD exceeds your spending need, we compute the marginal tax on the excess at your year-specific bracket position and reinvest the after-tax remainder into your brokerage account — preserving wealth that would otherwise be unnecessarily drawn down. This is the same approach used in professional planning software; most free calculators either skip RMDs entirely or treat the full RMD as consumption. After portfolio depletion, the projection shows what you can actually fund from remaining sources (guaranteed income), not your original goal — so the "spending" line reflects reality rather than wishful thinking.

Bridge reserve mechanics. The reserve can be funded from explicit cash savings (the "Cash reserve" input) OR earmarked from expected one-time inflows (home sale proceeds, inheritance, severance, other lump sums). Each source has its own toggle in Inputs — flipping the earmark routes that money to the bridge bucket instead of the general portfolio. The reserve grows at your high-yield savings rate (not equity returns), reflecting that it's parked in HYSA-style cash — protecting it from sequence-of-returns risk during the highest-risk withdrawal window. Bridge reserves not consumed during bridge years remain available as a last-resort source after all other buckets are exhausted (the engine tracks them through to end-of-plan rather than orphaning unused balance).

Research shows retirees spend more in early retirement (active years), less in the middle (slower pace), and more again late in life (healthcare). Every plan models three distinct spending phases — Go-Go, Slow-Go, and No-Go — each running independently through every simulation. You can adjust each phase amount and the age at which the next phase begins.

New plans default to Go-Go (100% of monthlyGoal), Slow-Go (85%), and No-Go (75%) — fully customizable in the Inputs tab. Setting all three phases equal models flat spending across retirement. monthlyGoal is a derived weighted average across the three phases, not a fixed monthly draw.

The model uses a single pre-retirement return rate and a single post-retirement return rate for your entire portfolio. This is a deliberate simplification — a full plan would model equity/bond allocation with correlated assets. For scenario modeling, the user controls these assumptions directly, which is more honest than hiding them inside an allocation model.

Healthcare costs inflate at the higher of your general inflation rate plus 2% or 5%, reflecting the historical premium of medical inflation over general CPI (typically 5-7%). This applies automatically — you don't need to enter an inflated healthcare estimate. If your actual healthcare spending will track general inflation more closely (some Medicare recipients), you can lower your monthly healthcare input directly. To test your plan's sensitivity to return assumptions, try running your scenario at 1–2% lower than your expected return.

Smart Moves are concrete actions you can apply to your plan to improve specific outcomes — earning a part-time bridge income, converting to Roth, delaying Social Security, harvesting capital gains, and others. The Results tab organizes them into four lenses, each measuring impact in the currency that matters to that goal: Retire Earlier (how much younger can you stop?), Spend More (how much more can you sustain?), Leave a Legacy (how much larger is the median ending balance?), and Weather a Downturn (how much higher is the survival rate under stress?). Click any move to see its individual impact in the active lens; toggle a move on to apply it to your plan and see the live effect on every other metric.

Each move's impact is measured against your CURRENT plan (baseline), not against other selected moves. The "See if this combined plan works" button at the bottom of each lens runs a unified Monte Carlo with all selected moves applied together — capturing interaction effects (e.g., part-time bridge income + Roth conversion timing + delayed SS) that aren't visible from individual impact deltas summed. Some moves are lens-aware: spend-reduction moves are excluded from the Spend More lens (contradicts intent); retirement-deferral moves are excluded from Retire Earlier (same reason); gain harvesting only appears when bridge years exist. The AI Advisor can recommend specific moves in conversation and deep-link you to the relevant lens with the move pre-toggled — useful when you want to explore a specific strategy without manually scanning all four lenses.

When multiple moves are selected, independent moves sum their deterministic deltas. For known interacting pairs — bridge income + Roth conversion, retire later + SS delay — the combined bar runs 1,000 Monte Carlo simulations to capture the joint effect accurately, since these moves meaningfully affect each other.

Single-move selections always use the card delta directly. The combined bar updates live as you select and deselect moves.

Ignores your entered retirement age. Scans candidate ages 50–75, runs 750 Monte Carlo simulations at each, and returns the earliest age that hits 85% confidence. When moves are applied, each move's override is recomputed at the candidate age — so bridge sizing, SS windows, and other age-dependent overrides are accurate for each age tested, not pre-baked from your entered retirement age.

750 sims per age balances scan speed with accuracy. The 85% threshold is fixed as a common benchmark. For very long retirements (40+ years), a higher target like 90–95% is more appropriate.

Scans Social Security claim ages 62–70 and finds the age that best fits your specific scenario. For early retirement plans, near-term income is weighted more heavily because the bridge period cost is real. For normal retirement timing, lifetime benefit optimization dominates.

This is a heuristic that finds a good answer for most scenarios — not a full actuarial NPV analysis. A rigorous breakeven calculation would account for joint life expectancy and mortality tables for married couples. For high-stakes SS timing decisions, specialized tools or a CFP are worth consulting.

We model three lifetime phases of building Roth assets: direct Roth IRA contributions while working, the Roth conversion sweet spot during your bridge years, and backdoor + mega backdoor Roth for high earners locked out of direct contributions. Each phase has different rules, eligibility, and tax treatment.

Backdoor and mega backdoor amounts are sourced from after-tax savings, so we deduct them from your brokerage contribution to avoid double-counting. The pro-rata rule applies if you have existing Traditional IRA balances — we surface this as an informational flag rather than modeling the tax cost precisely (would require knowing your trad IRA balance). When you set an annual Roth conversion amount, the engine routes that amount from 401k to Roth each bridge year (until the 401k is exhausted), computes the marginal income tax using your year-specific bracket position, and pays the tax from savings or brokerage. Conversions show up in your projection as a 401k → Roth movement and reduce future RMDs. The three phases address different lifetime windows: direct contributions during income years, conversions during low-income bridge years, and backdoor when you're income-gated out of direct contributions. LTCG bracket-cliff handling: when you have significant taxable brokerage drawing in the same year, the sweet spot calc also checks the LTCG 0%/15% bracket boundary — a conversion that pushes your ordinary income high enough to spill long-term capital gains from the 0% bracket into 15% is bracket-cliff territory, and the sweet spot conservatively backs off below that line. This prevents the sweet spot from recommending a conversion amount that triggers an unexpected 15% LTCG tax on your brokerage withdrawals in the same year.

Roth conversion ladder (FIRE framing). The FIRE community refers to a multi-year conversion sequence as a "ladder" — converting a deliberate amount each bridge year, then withdrawing seasoned conversions to fund pre-59½ retirement. Each year's conversion starts its own IRS 5-year clock for penalty-free principal access (separate from the contribution and earnings 5-year rules). The engine models the conversion side year-by-year — amount, tax impact, future-RMD reduction — and the sweet-spot calc above helps size each rung. The engine doesn't enforce the withdrawal-side 5-year clock, so anyone running a true ladder needs to track each conversion's seasoning themselves and pair the strategy with a bridge reserve to cover the first 5 years before any conversion is withdrawable.

The Compare tab includes a Median Saver in Your Cohort benchmark — automatically matched to your age band, income group, and household type. The benchmark uses cohort-appropriate balance, contribution rate, and Social Security benefit, but mirrors YOUR retirement goal and structural assumptions — so the comparison answers "if I had what a median saver has, would my plan still work?" rather than "can a median saver fund a synthetic 75% replacement target?" (the latter produces 0% success for most cohorts and isn't useful).

Sources: (1) Federal Reserve Board, "Changes in U.S. Family Finances from 2019 to 2022" (October 2023), Box 1 Table A — household-level mean retirement balances by income percentile group, working families ages 35-64 with retirement accounts (combined IRA + DC); (2) SSA Office of the Chief Actuary, Actuarial Note 2025.3 (June 2025) — hypothetical scaled worker AIMEs already account for real career arcs from the Continuous Work History Sample, with PIAs computed using 2024 bend points and the standard 90/32/15 progressive formula; (3) Center for Retirement Research at Boston College, Issue Brief 23-25 — published median 401(k)+IRA balances by age band for working households, used to calibrate the age-progression factors. The benchmark is anchored to published, household-level data — no per-individual synthesis. The mean→median conversion uses the ~2.0× ratio observed between SCF's published mean ($331k for 35-64 working savers) and CRR's published median ($204k for 55-64 working savers with a 401(k)).

Home sales, inheritances, and other expected lump sums each get different tax treatment based on real-world IRS rules. The amount that lands in your brokerage is the after-tax net, not the gross.

Home sale: We apply the IRS Section 121 primary residence exclusion ($250K single / $500K married) and tax the excess at 15% federal LTCG plus your entered state tax rate. The 15% federal rate is an approximation — your actual rate could be 0% (very low total income that year) or 20% (very high), but 15% is the standard rate for typical retirees. Most states (CA, NY, NJ, MA) tax LTCG at ordinary income rates, so applying your state rate to the gain is conservative and matches how most state tax codes work. Section 121 requires meeting the IRS ownership and use tests (lived in home as primary residence at least 2 of the last 5 years). If you live in a no-state-tax state (FL, TX, NV, WA), set state tax to 0% in your inputs and only the federal portion applies. Inheritance: Generally not subject to income tax under stepped-up basis rules — the recipient's cost basis is reset to the fair market value at the decedent's date of death, so liquidating immediately produces zero capital gain. We model inheritance as arriving untaxed. (Federal estate tax only applies to estates above $15M individual / $30M married in 2026; if you expect that, enter the after-estate-tax inheritance amount.) Other lump sum: Treated as a generic windfall routed at face value. If your expected lump is taxable (severance, deferred comp, lottery, business sale gain), enter the after-tax amount you expect to receive.

The Stress Test tab models specific, named risks rather than abstract market volatility. Each scenario card runs 1,000 Monte Carlo simulations with one stated change applied to your plan — nothing hidden. The SS Cut card models exactly your SS reduction; it does not also lower your return assumption or shorten your retirement. The combined bar runs a single MC pass with all selected moves applied jointly, so interacting effects (bridge income + Roth conversion timing) are captured accurately rather than approximated as a sum.

The stress band on the Projection chart shows what happens if you retire into a 5-year bear market (sequence-of-returns risk). It uses the same calculation engine as the deterministic projection and Monte Carlo — same per-account sequential drawdown, same tax math, same Roth conversion modeling — and only differs in applying a -8% mean return to the first 5 retirement years. For most people this is the most important stress test in the tool: a bad market in early retirement, when your portfolio is largest and you're drawing down, is far more dangerous than the same bad market 15 years in. Plans that look healthy under normal Monte Carlo can show dramatically lower survival rates under sequence-of-returns stress, especially if heavily weighted toward equities. If your stress band looks much worse than your Monte Carlo result, that is the model telling you something true about your specific plan — not a glitch. Common responses: hold more cash or short-term bonds at retirement, delay retirement by 1-2 years, reduce early-retirement spending, or build a larger bridge reserve.

Most calculators ask for a single state tax rate and apply it to every dollar of retirement income. State tax codes are far more varied than that. Pennsylvania doesn't tax retirement income at all. Hawaii excludes employer pensions but still taxes 401(k) withdrawals and Roth conversions. Five states tax Social Security at typical retirement incomes. Twenty-one states exclude part of retirement income once you reach a set age, usually up to a dollar cap and often only below an income limit — Georgia, for example, excludes up to $65,000 per person from 65, and New Jersey up to $100,000 for a couple under its income limits. Nine states apply preferential rates to long-term capital gains. We model 51 jurisdictions individually using their actual structure — when you pick your state in Inputs, the engine looks up the right rules and applies them year-by-year through your entire projection.

Concrete impact — a married couple with $1.5M nest egg, $7,500/mo goal, retiring at 65 ends with $8.05M in Florida (no state tax), $7.96M in Pennsylvania (excludes retirement income but taxes the small brokerage portion of withdrawals), $7.39M in New York (4.5% effective on retirement income), and $7.01M in Minnesota (6.09% rate plus Social Security taxation). Same plan, same dollars in, $1M+ different at life expectancy purely from where you live. Documented simplifications: progressive brackets are modeled as a single effective rate rather than the full graduated structure (this matters less in retirement, when most people fall into a narrow bracket band — the effective rate is calibrated to typical retirement-income levels); city and local taxes (NYC, San Francisco, Yonkers, Detroit) are not modeled, so urban residents in those cities should mentally add 1-3% to their effective rate; phased pension exclusions in Kentucky, New Jersey, New York, Virginia, and Maine are treated conservatively as "fully taxed" rather than partially excluded (so plans in those states are slightly under-projected, which biases toward caution); Washington's 7% capital gains tax above $278K is modeled at the LTCG rate but the threshold itself is not — most retirees won't cross it. Federal tax modeling is unchanged from Section 4. The state and federal pieces compose: federal first, then state on the relevant income depending on the state's rules. Source data is calibrated against the Tax Foundation's State Individual Income Tax Rates report, the Kiplinger State-by-State Guide to Taxes on Retirees, and each state's revenue department for the SS-taxation and pension-exclusion specifics.

During your bridge years, if your taxable income drops low enough, long-term capital gains can be "harvested" at 0% federal tax — sell appreciated brokerage holdings and immediately rebuy them to reset your cost basis upward, locking in the gain at zero tax cost. This eliminates future tax on those harvested gains. The "Harvest gains" Smart Move enables this in your projection.

The 0% LTCG bracket is one of the cleanest tax wins in retirement planning, but it's only available when your other taxable income is low — bridge years before Social Security starts. When the harvest move is enabled, the engine each year checks: (1) your projected taxable income for the year (post-deduction); (2) the remaining headroom under the 0% LTCG bracket cap; (3) your unrealized brokerage gain (balance − basis). Whatever fits gets harvested — basis steps up, no federal tax owed on the harvested portion. State tax may still apply depending on your state (most states tax LTCG at ordinary rates without a preferential 0% bracket; the eight that DO have preferential LTCG rates are honored). The move only fires when there's meaningful headroom and meaningful gain to harvest — it doesn't bother with sub-threshold amounts where the bookkeeping wouldn't change the projection. Pairs well with Roth conversions in the same bridge years: both strategies use the same low-income window to convert future tax burden into present-zero tax. The conversion sweet spot calculator (Section 15) and the harvest move are complementary — the sweet spot stops before pushing LTCG into the 15% bracket, leaving room for the harvest move to fill the rest of the 0% bracket.

An opt-in dynamic spending strategy that adjusts annual spending based on portfolio performance instead of using a fixed inflation-adjusted withdrawal. The strategy applies two rules: "capital preservation" cuts spending when your withdrawal rate climbs too high in a bad market; "prosperity boost" bumps spending when your rate drops too low after a good run. The academic research finds initial withdrawal rates of 5–5.5% historically sustained under this framework — meaningfully higher than the static 4% rule. Guardrails are mutually exclusive with phased spending (Go-Go / Slow-Go / No-Go): when enabled, phased values are paused and the engine uses your monthly goal × 12 as the year-1 baseline.

Year-1 spending is always the baseline (your monthly goal × 12 × inflation). Rules fire starting year 2 based on the prior year's actual withdrawal rate. The investable portfolio (for the WR denominator) excludes bridge funds (earmarked home-sale net, inheritances, other lumps) because those are pre-allocated for the bridge period and not part of the long-term investment portfolio the strategy is meant to govern. Strict interpretation of the paper applies the rules throughout retirement, including bridge years — even though bridge years naturally have a higher withdrawal rate. The paper's wide band (±20%) accommodates this, but expect more frequent rule triggers during bridge years for early retirees. SMART MOVES CAP — when Guyton-Klinger is applied as a Smart Move (Spend More or Retire Earlier lens), the engine caps the headline at a 5.0% initial withdrawal rate. The Guyton + Klinger 2006 paper's empirical safe-start range is 4.5–5.5%; we anchor at the conservative end. Subsequent research (Pfau 2010, Kitces multiple) has questioned whether the upper end of the historical safe-start range remains defensible in lower-yield regimes — the conservative cap respects that critique. Without this cap, the finder algorithms could converge on inflated headlines that the engine technically sustains via aggressive preservation cuts but which misrepresent what the paper actually backs. For the Retire Earlier lens specifically, the engine additionally requires that the strategy's average lived spending across surviving Monte Carlo simulations meets or exceeds the user's stated monthly goal — a calculator design choice (not paper-derived) to ensure "you can retire at age X" implies "spending what you said you wanted to spend." LIVED EXPERIENCE UNDER GUARDRAILS (Results card) — when Dynamic Spending is active, the Results tab surfaces a dedicated card showing what the strategy actually looks like across simulated retirements. Three-tier severity on the "Years below target" tile (default text under 30%, amber 30–50%, red ≥50%) gives an at-a-glance read on how often spending falls below your stated monthly goal. The "Worst-10% spending floor" tile surfaces a concrete dollar gap and percentage below your goal — making the floor's lived meaning legible at a glance. A bad-sim narrative above the spending trajectory chart names the actual cut-fire ages from your worst-lived-experience surviving simulation ("In a tough sequence, cuts fired at ages 61, 62, 63, 64, and 68 — 7 years below your stated goal, with 4 of those consecutive"). Picks the surviving sim with the most below-target years so you read a real lived-experience story, not a worst-case hypothetical. Empirical research expects ~1–1.5pp success-rate uplift from the strategy on typical plans. The FIRE community references this strategy as "Guyton's guardrails" or "variable percentage withdrawal." Reference: Guyton, J. T., and Klinger, W. J. (2006). "Decision Rules and Maximum Initial Withdrawal Rates." Journal of Financial Planning. See also Pfau (2010) "Safe Savings Rates" and Kitces (multiple articles) for the post-2008 lower-yield critique. RESEARCH-COMMUNITY CRITIQUE (calculator's honest framing) — the Guyton + Klinger 2006 paper calibrated against historical data through ~2005, when stock valuations (Shiller's CAPE / cyclically-adjusted P/E ratio) and bond yields sat in a meaningfully different regime from today's market. Subsequent research has raised valuation-based concerns about whether those historical safe-start rates transfer cleanly to high-CAPE starting environments: Pfau (2010, 2012, multiple) finds that high-CAPE start years materially compress safe withdrawal rates across historical replays; Karsten Jeske's "Big ERN" Safe Withdrawal Rate series (2017+) extends the analysis to longer (40–60 year) retirements common in FIRE planning and concludes the historical 4–5% rules-of-thumb may overstate safety for early-retiree windows. The 5.0% Smart Moves cap respects this critique by anchoring to the conservative end of the 2006 paper's 4.5–5.5% range. Users with sub-3.5% effective withdrawal rates have substantial buffer regardless of this debate; users at 4–5% are operating in the contested window where the GK strategy's historical defensibility is genuinely debated in current academic literature. We surface the rules and the lived-experience tradeoff; we are not committing a position on whether GK guardrails will provide the same 1–1.5pp uplift forward as they did historically. WHAT THE CALCULATOR DOES AND DOESN'T DO WITH CAPE — your specific start-year CAPE (Shiller's cyclically-adjusted P/E ratio) is the empirical variable the Pfau + Big ERN bodies of work argue determines whether the 2006 paper's range is conservative or aggressive at your retirement year. The engine does not read your start-year CAPE and does not fold it into any recommendation — the 2006 paper's empirical 4.5–5.5% range is the calibration; the 5.0% cap is the conservative-end anchor that hedges against this exact uncertainty. We deliberately don't take a position on whether today is a high-CAPE environment because that reading changes year-over-year, requires picking + citing a specific CAPE source, and the conservative-end anchor's defensibility doesn't depend on the answer. If you want to interpret the cap against today's valuation environment, the cited Pfau and Big ERN work is the place to start that reading.

An opt-in workflow for users who plan to save aggressively until their existing balances will compound to their retirement target — then stop contributing and "coast" (keep working but with no new savings going in) the rest of the way to traditional retirement. The math is the same compound-growth formula running all retirement projections; the workflow change is splitting the future-value calculation into two phases: a contribution phase (current age → coast-end age), then a growth-only phase (coast-end age → retirement age).

When Coast FIRE is active for a party, the engine stops ALL retirement contributions for that party at the coast-end age — 401(k), Roth IRA, brokerage, HSA, and the employer match (employer match is contribution-linked, so it ends when contributions end). Catch-up contributions still apply DURING the contribution window if the party is 50+ (the wrapper passes the truncated window to fvWithCatchup, which handles catch-up logic internally). The defaults are zero — Coast is opt-in. Most users will keep contributionEndAge = 0 (and spouseContributionEndAge = 0), which preserves the conventional "contribute right up to retirement" behavior with byte-identical engine output. Out-of-range values are handled gracefully: a coast-end age before currentAge collapses to "no contributions from now" (just compound existing balances); a coast-end age at or past retirementAge reduces to sentinel behavior. SMART MOVES INTERACTION — when Coast is active, the Smart Moves workshop filters contribution-boost moves (max 401(k), max Roth, employer match, catch-up routing, backdoor Roth, mega backdoor Roth, HSA contributions, and spouse equivalents) from the applicable lists across all four lenses. They don't make sense as recommendations when the user has explicitly opted to stop contributing. Non-contribution moves (SS timing, Roth conversions, spending discipline, gain harvesting) still apply normally. WHAT THIS DOESN'T MODEL — Coast FIRE changes what happens to your retirement accounts, not your income or your spending. The engine assumes you're still working through retirement age (your wages cover expenses); it just doesn't direct any portion of those wages into retirement accounts after the coast-end age. If your real-world Coast involves a meaningful salary cut, edit yourSalary directly to reflect it. WHY THIS WORKFLOW — Coast FIRE is widely discussed in the FIRE community as a less-extreme version of FIRE: instead of front-loading enough savings to retire at 45, you front-load enough to remove savings pressure by 45 and let compound growth carry you to 60–65. The workload change matters more than the timeline change — Coast lets people downshift to lower-paying, more enjoyable work without retirement-math anxiety.

An input for modeling part-time work income during the retirement window — the FIRE variant where retiring from a career doesn't mean retiring from all wage income. You enter an annual wage amount and the age range it covers (start age → end age, both inside your retirement window). During those years, the wage reduces the portfolio withdrawal need and stacks on top of any other ordinary income for tax purposes.

The interactions that matter for FIRE planning. Tax bracket. Barista wage is ordinary income and stacks with portfolio draws and SS — a high enough wage can push you into a higher bracket and squeeze the Roth conversion sweet spot (Section 15) during the same years. Social Security earnings test. If you claim SS before your Full Retirement Age and your barista wage exceeds the annual earnings limit (~$22,320 in 2025, inflation-adjusted forward), the SSA reduces your benefit $1 for every $2 over the limit. If this applies to your plan, model the reduced benefit by lowering the "SS Benefit Confidence" slider accordingly. IRMAA. Barista wages count toward MAGI when 65+, so they can push you across an IRMAA tier and add Medicare premium surcharges (Section 4 covers the bracket mechanics). ACA subsidies. Pre-65, MAGI determines marketplace subsidy eligibility — a barista wage that crosses the cliff can substantially raise out-of-pocket healthcare cost. What this doesn't model. FICA tax on the wage itself (~7.65% W-2 / 15.3% self-employed) is not subtracted from the wage you enter — use your expected after-FICA take-home if precision matters. Employer 401(k) match during barista years isn't separately modeled; if your barista job offers a match, fold it into your savings rate. Why this workflow. The FIRE community widely treats Barista FIRE as a hedge — partial wage income reduces sequence-of-returns risk in the early retirement window and lets people retire from their career years earlier than full-FIRE math requires.

The window between retirement and Medicare eligibility (typically retirement age → 65) is one of the most under-modeled costs in early retirement. Employer coverage ends, COBRA expires after 18 months, and an ACA marketplace plan becomes the realistic option for most households — at premiums driven by household MAGI relative to the federal poverty level. The engine treats this bridge as a first-class input rather than rolling it into general spending, because the dollar exposure is large ($15k–$30k+/yr per adult is typical for full-cost marketplace plans without subsidies) and the MAGI interaction creates real planning leverage that disappears once Medicare starts.

The mechanics that matter across the bridge. Pre-65 cost input. Monthly healthcare goes into Inputs as a dollar figure; the engine inflates it year-over-year at the higher of general inflation + 2% or 5% (medical inflation has historically run 5–7%, well above CPI). The runway from today to retirement is fully inflated, not just one year forward — pre-fix (May 2026), plans with long timelines under-stated pre-Medicare cost by 50%+. ACA subsidy and the cliff. ACA marketplace premium tax credits scale with household MAGI; the engine tracks the per-state cliff thresholds and refreshes them annually. The Cliff Proximity Gauge on the Roth Strategies card shows live MAGI position vs. the ACA cliff and the five IRMAA tiers — so you can see whether a Roth conversion or gain-harvest in a given bridge year pushes you across. IRMAA at 65+. Once a household member is Medicare-eligible, MAGI above the tier-1 threshold triggers Part B + Part D surcharges, applied per person crossing the threshold (Section 4 covers the tier-by-tier mechanics; thresholds inflate at CPI to track real-world bracket creep). The bridge planning question. Which bracket-management strategies — Roth conversions (Section 15), gain harvesting (Section 20) — make sense in the bridge years? Both lower future RMDs and IRMAA exposure, but both raise current MAGI and can shrink ACA subsidies if you're still pre-65. The Cliff Proximity Gauge exists for exactly this tradeoff. What this doesn't model. COBRA (the 18-month employer-plan continuation) isn't a distinct path — if you plan to use COBRA before the marketplace, enter the COBRA premium for those years. State Medicaid expansion variations (eligibility at low MAGI differs by state) aren't modeled. Long-term care premiums are modeled separately under Inputs → Life Events → Long-Term Care, not in this healthcare-bridge mechanic. Medigap and Part D plan selection aren't optimized — the IRMAA surcharge applies on top of whatever baseline Medicare premium you've factored into your monthly input. Why this matters. Pre-65 healthcare is one of the most-cited reasons FIRE-leaning households delay retirement, and the most mis-estimated line item in early-retirement budgets. Modeling it explicitly — with the ACA cliff and IRMAA cliff visible together — turns "is healthcare going to wreck my plan?" into a question the math can actually answer.

Replays your plan against the ACTUAL year-by-year market returns and inflation from a past retirement-start year, drawn from Robert Shiller's canonical 1928–2022 dataset (S&P 500 nominal total return, 10-year Treasury nominal total return, CPI-U inflation). This is the Bengen (1994) and Trinity Study (1998) methodology that originated modern withdrawal-rate research — decades before stochastic Monte Carlo was widely available. Each year’s spending comes out at the start of the year and what is left earns that year’s return, the convention used by Bengen, the Trinity Study, FIcalc and cFIREsim. (The Monte Carlo and the projection take spending at the end of the year, after that year’s growth.) Surfaced two ways on the Stress Test tab: a Historical Back-Test card (one of five named eras, cohort-survival ring) and the Historical Robustness Workshop (full-range exploration, every eligible start year + stock allocation).

COHORT SR vs MONTE CARLO SR — Both express probability on the 0–100 scale, but they measure different sample spaces. Monte Carlo SR is the fraction of 1,000 stochastic-sample futures where the plan succeeds (random draws from a normal distribution). Cohort SR is the fraction of historical start-year cohorts where the plan would have survived. Neither is "more correct." Monte Carlo explores tails that never historically occurred (its left tail is wider than history); historical replay captures mean reversion that the MC model doesn't have access to. How close the two land depends mostly on the return you assume: if your assumption is lower than what history actually delivered, the replay will look better than the simulation, and the reverse if it is higher. For thin-margin plans (4%+ WR, 30+ yr retirement) historical replay is where dramatic divergences appear: Bengen famously showed that the 1966–1973 cohort broke at 5% WR while later cohorts survived at 6%. METHODOLOGY COMPLETENESS — Investment buckets (401(k), Roth, Brokerage, HSA) grow at the historical year's actual market return. Spending goal, Social Security, pensions with COLA all scale via historical CPI-U. Tax brackets (IRMAA, LTCG) inflate via the historical sequence × pre-retirement constant rate. HYSA + Bridge Cash buckets grow at the historical year's inflation rate (real return ~0% — matches Shiller's long-run T-bill empirical regularity; floored at 0% nominal to prevent shrinkage during deflation). Slightly conservative in eras where T-bills genuinely paid above inflation (e.g., 1980s) but unbiased on average. This closes the methodology seam that previously existed where cash buckets grew at the user's modern HYSA assumption regardless of era. ERA_WINDOWS CURATION — Five canonical eras, each a rolling window bracketing the named stress event with starts on either side: Great Depression 1929 (window 1928–1933, 6 starts), 1966 Bear Market (1965–1969, 5 starts), Stagflation 1973 (1969–1976, 8 starts — the canonical FIRE break window), Dot-Com Bust 2000 (1999–2003, 5 starts), Financial Crisis 2008 (2007–2010, 4 starts). References: Bengen, W. P. (1994). "Determining Withdrawal Rates Using Historical Data." Journal of Financial Planning. Cooley, P. L., Hubbard, C. M., & Walz, D. T. (1998). "Retirement Spending: Choosing a Sustainable Withdrawal Rate" (the Trinity Study). Shiller dataset: http://www.econ.yale.edu/~shiller/data/ie_data.xls.

Your monthly goal handles ongoing living costs — housing, food, transport, healthcare, leisure. Future Expenses is a separate list for the lumpier outflows that arrive at specific ages: a daughter's wedding at 68, four years of college help from 58, a kitchen remodel at 75, a car loan's payments to 72. You enter each in today's dollars and the engine inflates it to each year it fires — or mark it "Fixed amount" and the typed figure is charged exactly, never inflated: the right treatment for a contract price (a loan payment, a lease). An entry can be one-time or repeat — every year through a chosen age (a payment: tuition, a loan), or every few years (a repeating lump sum: a new car every 7 years, a roof every 15). Up to 15 entries. Ongoing lifestyle costs still belong in your monthly goal.

The cascade reuses the same withdrawal order as the home purchase deduction (§17 Lump Sum Tax Handling) — savings first (no tax friction), then brokerage with basis-adjusted gain tracking (the engine applies the same _drawBrok cascade helper used everywhere else), then Roth (no tax), then 401(k). This isn't tax-perfect for every individual situation (tax on the 401(k) portion would be assessed in the year of draw at your marginal rate), but it produces the structurally correct cash flow and follows the same conservative ordering every other lump-sum deduction uses. UI-MANAGED ONLY — Future Expenses is the only DEFAULTS field in the app that the AI Advisor explicitly cannot modify via APPLY. The AI has full READ access to your list (it can reference "your $30k wedding at 68" in narrative) but cannot add, edit, or remove entries on your behalf. The list is yours: you stay in control of what counts as a planned one-time expense and what doesn't. This is enforced two ways: server-side (futureExpenses is intentionally excluded from the AI's fields[] catalog in the chat prompt) and client-side (the applyToPlan central pipeline drops any futureExpenses field from inbound APPLY changes before committing). If you want the AI's help thinking about a planned expense, ask it conversationally; the AI will discuss the trade-off and point you to the Future Expenses panel in Inputs → Life Events to add the entry yourself. WHAT THIS DOESN'T MODEL — Ongoing LIFESTYLE costs (annual vacations, ongoing parental support, predictable medical co-pays) still belong in your monthly goal — use spending phases for those. But a FIXED-TERM recurring cost (college tuition over four years) is a PAYMENT — entered once as an amount per year with a last-payment age — and a purchase that comes back (a new car every 7 years) is a REPEATING LUMP SUM, entered once with its gap; the engine draws each occurrence in the year it fires, priced in that year's dollars. Expenses that fund through specific accounts (HSA-only medical, 529-routed tuition) are simplified to the standard cascade; the engine does not pre-allocate from a designated source. Since August 2026 the 401(k) portion of a draw is charged its real tax in the year of the draw — the engine withdraws enough extra to cover the tax bill itself. The list cap is 15 entries (raised from 10 in August 2026); that's an architectural limit, not a research recommendation — most retirement plans don't need more than 5–10 entries to capture material outflows. HOW TO USE IT WELL — Estimate in today's dollars; let the engine handle inflation. Be conservative on amount: $35k for a wedding rather than $25k unless you really have it dialed in. Be specific on age: "sometime in my 70s" isn't actionable — pick a specific year. Run Monte Carlo before and after adding expenses to see how your plan absorbs them — if a $30k expense at 70 visibly drops your success rate, that's the model telling you the expense is straining your plan and you may want to budget for it in advance (more savings now, smaller monthly goal in retirement, or accepting the trade-off honestly).

A Qualified Charitable Distribution is a direct gift from a Traditional IRA to a qualified charity, allowed once you're 70½. Unlike a normal charitable deduction, a QCD is EXCLUDED from your income entirely — it never appears as taxable income in the first place. It counts toward your Required Minimum Distribution tax-free, and because it keeps your income (and therefore your MAGI) lower, less of your Social Security is taxed and you stay further from the Medicare IRMAA surcharge cliffs. The honest trade-off: the donated principal leaves your estate, so it lowers your legacy. Two inputs in the Tax section — a fixed annual amount, or “give my entire RMD to charity” (which tracks the RMD as it grows year over year).

THE HEADLINE CHANNEL IS IRMAA — a normal RMD inflates your MAGI; a QCD-satisfied RMD does not. For a retiree sitting just over an IRMAA tier threshold, redirecting the RMD to charity can drop MAGI back under the cliff and erase the Medicare surcharge for both spouses. The SS-taxability and federal-bracket savings fall out automatically as the taxable RMD shrinks. IRA-ONLY, AGAINST A POOLED BUCKET — a QCD must come from an IRA, not a 401(k). The engine pools all pre-tax balances (401(k) + Traditional IRA) into a single bucket, so it models the QCD against that pooled pre-tax balance; if your pre-tax money is mostly in a 401(k), roll the portion you intend to give into an IRA first (a standard tax-free rollover) so the real-world mechanics match the model. USER-ONLY (v1) — the QCD applies to your own pre-tax bucket; because the pooled bucket cannot attribute whose IRA a dollar came from, a separate spouse QCD is not modeled in this version. WHAT THIS DOESN'T MODEL — the 70½ half-year is applied from age 70 (whole-year engine). Full-RMD mode is the precise, correctly-timed path (it auto-starts at your RMD age); a fixed annual QCD applies from 70 regardless. State tax follows your state's general retirement-income rules. BYTE-IDENTICAL WHEN OFF — with no QCD set, every QCD computation reduces to zero, so plans without a QCD are unchanged. References: IRS Publication 590-B; Internal Revenue Code §408(d)(8).

Withdrawals from pre-tax retirement accounts (401(k), Traditional IRA) before age 59½ normally incur a 10% federal early-withdrawal penalty on top of ordinary income tax. The engine applies this penalty by default. Because the engine draws taxable accounts first (bridge reserve → savings → brokerage → 401(k)/IRA → Roth), only plans that exhaust their taxable accounts before 59½ ever reach the penalized pre-tax bucket — in practice, marginal early-retirement (FIRE) plans. A toggle in the Tax section (“penalty-free access before 59½”) turns the penalty off for users with a qualified early-access strategy: a 72(t)/SEPP schedule, the Rule of 55, or a Roth conversion ladder.

DEFAULT-ON IS THE HONEST FLOOR — modeling the penalty by default means an aggressive early-retirement plan shows its true cost: a 50-year-old draining a 401(k) to bridge to 59½ loses 10% of every pre-tax dollar to the IRS. Turning the toggle on is the user asserting they have a real penalty-free mechanism in place. WHO ACTUALLY HITS IT — most plans don't, because the engine spends taxable money first; the penalty only bites when taxable + Roth-contribution money run out before 59½. That's exactly the thin-margin FIRE population for whom the 10% matters most, and the population most likely to have a 72(t) or Rule-of-55 in place — hence the toggle. WHAT THIS DOESN'T MODEL — the penalty on Roth EARNINGS withdrawn before 59½ (would require Roth cost-basis tracking the engine doesn't carry; Roth is drawn last anyway); per-spouse age within the combined household pre-tax bucket (uses the primary age); state-level early-withdrawal penalties (federal 10% only). The toggle trusts your assertion — it does not validate that you've set up a compliant 72(t), satisfied the Rule of 55's separation requirement, or seasoned a ladder five years. Reference: Internal Revenue Code §72(t).

If you claim Social Security before your Full Retirement Age (67 for anyone born in 1960 or later) and you keep earning a paycheck, the SSA applies an “earnings test”: it withholds $1 of benefits for every $2 you earn above an annual limit (~$24,480 today). The withheld benefits aren't lost — at your FRA the SSA recomputes your benefit upward to credit back the withheld months, so over a normal lifespan it's roughly actuarially neutral. The real cost is timing: less SS during the early working years (when your portfolio is largest and sequence risk is highest), more later. It's the classic Barista FIRE surprise. The engine applies this automatically — there's no toggle; it triggers whenever you claim before FRA and have part-time income above the limit.

ONE ADJUSTMENT, TWO EFFECTS — pushing the effective claim age later captures both halves in a single number: a LONGER BRIDGE (SS starts later → your portfolio funds more of the early working years) AND a HIGHER RESTORED BENEFIT (less early-claim reduction once it starts). Because the headline numbers, the year-by-year projection, and every Monte Carlo path read the same effective claim age, the three surfaces stay perfectly consistent. INFLATION-INVARIANT — earnings and the limit inflate together, so the withheld fraction doesn't depend on the inflation assumption; the calculation is done in today's dollars. INDIVIDUAL TEST — per-person, on your own SS based on your part-time income only; there's no spouse part-time-income field. WHAT THIS DOESN'T MODEL — partial withholding is modeled as deferral (a slight conservative bias); the higher special $1-per-$3 limit in the calendar year you reach FRA is not modeled (standard under-FRA $1-per-$2 throughout); only EARNED income counts (pensions, account withdrawals, investment income don't trigger it). When it applies, a note under your SS claim age shows the estimated annual withholding and your effective claim age. Reference: Social Security Administration, Retirement Earnings Test.

How long should your money last? Most people plan to their life expectancy — but that's a coin flip: by definition, half of people outlive it. Planning to the average means a 50% chance of running short in your final years, the single most avoidable retirement mistake. The Planning Horizon Workshop suggests a horizon you're unlikely to outlive by reading the actual survival curve, not the average — drawn from the Social Security Administration's 2022 Period Life Table (the same actuarial data Social Security uses). For couples it plans to the LAST survivor, which is meaningfully longer than either spouse alone.

PLAN TO THE TAIL — mean life expectancy is the wrong target: it's a 50/50 bet, and outliving your money is far worse than dying with a cushion. A percentile slider lets you choose your own margin (the default plans to roughly the 85th percentile — only about a 1-in-7 chance of outliving it); the mean is always shown alongside as the coin-flip for contrast. Optionally provide sex at birth for a sharper read — it's a workshop-only choice, never saved. FREE — the workshop and a gentle plan-health nudge (shown when your set life expectancy carries a real chance of being outlived) are free, not a paid feature. CAVEATS — the 2022 period table slightly understates future longevity (a conservative direction), and it's population-average, age + sex only: affluent, healthy, or long-lived-family profiles run longer, so lean toward the cautious end. Reference: Social Security Administration, Period Life Table, 2022 (2025 Trustees Report), table 4.C6.

The free calculator answers what your plan does. Deep Search answers which strategy is best — it puts the same simulation engine to work, running hundreds of scenarios against your actual plan to find the strongest choice for one decision at a time, judged on your real outcome — median legacy in today's dollars for the legacy-seeking tools (for Roth conversions, legacy after the income tax your heirs would owe on inherited pre-tax money), success rate for the risk-side scan. Ten searched optimizers: the couple's Social Security claim (all 81 combinations, survivor benefit included), the Roth Conversion Planner (formerly the Bridge Optimizer — conversion ladders across every year from retirement until the year before your first required withdrawal, plus shorter ladders that stop when Social Security starts, each tested with and without 0% gain harvesting, ACA subsidies and IRMAA cliffs priced into every candidate; scored on what your family keeps after the heirs' income tax, at the heirs' bracket you choose — charity, lower, middle or higher — with the answer shown at all four, and a materiality floor so a noise-level win never dethrones your current plan), what your plan hinges on (a tornado ranking which assumption you're most fragile to), solve-for-your-goal (name a target like 95% success and it works backward to the exact lever value that hits it), Spending Shape (the permission-to-spend search — the most you can spend, especially in your go-go years, while your odds hold a floor you choose and your legacy stays above what you want left, which can be zero), the Combined Risk Scan (the risk-side sibling — it stacks validated pairs of the Stress Test tab's one-at-a-time risks, reports honestly whether any combination hits harder together than the two cards predict separately, and when one does, prices the defense — dynamic spending guardrails — against your own worst pair), Employment Pause (how long you could stop working and still land the plan), Your Next Home (the whole housing timeline priced — sell, buy, rent, or finance), Spending Guardrails (the portfolio lines where you would trim or could raise), and Deep Scan (the one-click capstone that runs the legacy-seeking optimizers together for the true combined payoff — deliberately excluding Spending Shape, because spending more is a values trade, not a free win). It's the searched optimization the $5,000–$10,000 advisor tools charge for, opt-in behind a Run button in the Optimize tab, part of Navigator.

SEARCHED ON YOUR REAL PLAN — every candidate is a full run of the same Monte Carlo engine behind your headline numbers, so it reflects your actual taxes, bridge years, and sequence risk, not a rule of thumb. THE MOVES INTERACT — Deep Scan runs one combined simulation instead of adding up the parts, because delaying Social Security reshapes the room for Roth conversions; the true joint payoff isn't the sum. MONTE CARLO NOISE — under simulation variance, sub-dollar precision is illusory, so Solve-for-Goal grids and interpolates; searches run at a lighter sim count, then the winner is re-scored at the full display count. NO ENGINE CHANGE — Deep Search re-applies inputs the engine already reads, so it never alters how your plan is computed or your saved AI Plan Score. OPT-IN AND PAID — each optimizer runs only when you press Run (Couples SS alone is 81 full plans); it's part of Navigator. Reference: RetirementScenario.com, Optimize tab.

Most calculators assume you contribute the same dollar amount every year until retirement — but real savers raise their contributions as their income grows. Contribution escalation models that: pick any account (401(k), Roth, savings, brokerage, HSA — for you or your spouse) and set it to climb each year, either by a percentage that tracks your raises or a flat dollar step. Tax-advantaged accounts ramp up to the IRS limit, then hold. The result is a more realistic accumulation path — and a "Building your nest egg" table on the Projection tab that shows, year by year, how your contributions and growth compound into your retirement-day nest egg. There's also a free Smart Move that recommends ramping your 401(k) for savers who can't max out today but have the runway to grow into it.

RAMPS TO THE CAP, THEN HOLDS — for 401(k) and Roth, the yearly increase stops at the IRS contribution limit; it never assumes you'd contribute above the legal maximum. BOTH SPOUSES — each account, for you and your spouse, escalates independently. BYTE-IDENTICAL WHEN OFF — a plan with no escalation set computes exactly as before; the ramp only changes the math when you turn it on, so your saved AI Plan Score is unaffected until then. THE ACCUMULATION TABLE RECONCILES — every row reads Start + Contributions + Growth = End, and the final row equals the nest egg your retirement projection starts from, to the dollar, so you can validate it yourself. A MODELING ASSUMPTION, NOT A COMMITMENT — you tune it or turn it off anytime; real-world raises aren't guaranteed, so model conservatively. Reference: RetirementScenario.com, Inputs tab.