Answer: b. Everyone fears the crashes, and the crashes are the wrong villains. Black Monday 1987 was recovered in about two years; 2008 was savage but comparatively brief; even 1929 was a bad decade for everyone, prices included. What savaged the class of 1966 was the slow grind that followed: markets going sideways while the 1970s inflation quietly repriced everything they needed to buy. If you were working then, you got raises. A retiree drawing down savings got no raise, just a bigger grocery bill against the same pot. A crash is a fright; inflation against a fixed pot is a leak below the waterline.

This is why Part 1 of the app works the way it does. It doesn't invent market returns from a tidy formula. It replays your retirement through 10,000 alternative histories resampled from the real record — US stocks, bonds, bills and inflation, 1928 through 2025 — and it grabs whole blocks of five to ten consecutive years at a time, keeping each year's returns and its inflation together. That matters: bad years cluster, and the inflation that erodes your spending arrives in the same breath as the markets that bruise your bonds. Sample the years independently and you launder the 1970s into something polite. There are also two named stress tests — retirement beginning in 1966, and in 2000 — because "the worst starting years actually on record" is more honest than a made-up doomsday.

You tell the app five things: who's in the household, what income arrives monthly, what you spend, what you've saved, and how large a temporary loss you could stomach — in dollars, because "a 30% drawdown" is abstract and a dollar figure is not. The one idea I'd defend above all others: spending goes in three buckets — essential (money you cannot comfortably cut), lifestyle, and legacy. When money runs tight in a simulation, the app trims legacy first, then lifestyle by a slider you control, and never silently trims the essentials. An essential shortfall gets reported, loudly, with its timing: "in the model's difficult scenarios, essential spending first fell short around age 93." That's a sentence a couple can act on — trim the travel, not the groceries. And the whole thing is deterministic: same inputs, byte-identical results, every time. Your plan saves to a small file you keep; the app deliberately remembers nothing.

Answer: c, and this one offends people, because they've done nothing exotic. The culprit is what planners call the tax torpedo: each extra dollar of income can drag more of your Social Security benefit into taxability, so one dollar of withdrawal is taxed — and simultaneously causes previously untaxed benefit dollars to be taxed too. My test fixtures show a couple paying an effective 22.2% on the margin while sitting inside the "12%" bracket. The thresholds that govern this were frozen in nominal dollars back in the 1980s and '90s and have never been indexed, which is why a little more of everyone's benefit slides into the tax net each year without any law changing. The app doesn't estimate your marginal rate from a table — it literally recomputes your entire tax return with $1,000 more income and measures the damage. (And if you guessed d, you weren't fantasising — 40.7% is the same torpedo one bracket up, 1.85 × 22, for couples in the 22% bracket. Ours is in the 12%, where it tops out at 1.85 × 12 = 22.2%.)

Part 2 is the tax layer. Every constant in it — brackets, deductions, RMD ages, the temporary $6,000 senior deduction with its phase-out and 2028 expiry — was verified against the official 2026 sources, and each value carries its source and effective date inside the app. The engine knows which values grow with inflation and which are frozen, and it labels future years as what they are: projections under stated assumptions, not promises about future law. It also solves a genuinely circular problem properly: to raise cash you withdraw from an IRA, but the withdrawal is itself taxed, so you must withdraw more, which is taxed more… the engine runs a bisection search until the gap is under one dollar.

What you get from all this machinery is comparisons you can understand: which accounts to spend from first, and whether a Roth conversion schedule — fill the 12% bracket, fill the 22%, stay under the Medicare cliff, or name your own figure — improves your lifetime picture. The app will never bark "you should convert $80,000." It says favourable or unfavourable under these assumptions, which is the only honest phrasing there is.

Answer: d. Medicare premiums have income surcharges — IRMAA — and they work as cliffs, not slopes. At $218,000 of joint income you pay the standard premium. At $218,001 both spouses step up a tier. The tempting wrong answers are partial truths: about $96 is what one spouse pays in one month (an extra $81.20 for Part B plus $14.50 for Part D); $1,148 is one spouse's full year. But the surcharge applies to each Medicare enrollee, so the household pays about $2,297 a year. One dollar of income, four figures of premiums. And here's the trap within the trap: the surcharge arrives two years later, because Medicare looks at your tax return from two years back. The Roth conversion that feels clever in 2026 can ambush your premiums in 2028, after you've forgotten all about it. The app models the two-year lookback faithfully and warns you about a cliff two years before it lands — while you can still do something about it. This, more than anything, is why "just convert to the top of the bracket" advice from a golf partner needs checking against the actual cliffs.

Answer: c, and almost nobody believes it until they see the arithmetic. Claim at 62 (with a full retirement age of 67) and you keep 70% of your full benefit, forever. Wait until 70 and you get 124% — delayed credits add two-thirds of a percent for every month you hold out. (If you picked d, you half-remembered the right number: 124% is what you receive at 70, measured against your full benefit — it's the comparison against 62's 70% that makes the gap.) 124 against 70 is a 77% larger check, inflation-adjusted, for life. That doesn't automatically make waiting right — you're spending savings in the meantime, and the break-even depends on how long you live. Which is precisely the kind of trade a model should show you rather than a rule of thumb should decide for you.

The app's benefit engine uses the real SSA rules by the month — early-claiming reductions, delayed credits, spousal top-ups with deemed filing, the earnings test if you claim while still working — all verified against SSA's published factor tables. It compares a small, understandable menu: both at 62, both at full retirement age, both at 70, or the higher earner waiting to 70 while the other claims earlier. And if you've already claimed, it respects reality: it uses your actual benefit and refuses to "optimise" a decision you can't unmake. Lifespans, meanwhile, aren't a slider set to 90. Each spouse's length of life is drawn separately, thousands of times, from the official SSA life table — which says an average 65-year-old woman has a median age at death around 86–87, and a 4.6% chance of reaching 99. That one-in-twenty tail is why "plan to 85" is quietly dangerous, and why the app includes explicit long-life checks at 95, 100 and 105.

Answer: a. The smaller check simply stops. (The $3,300 of c is real arithmetic from the wrong chapter — "half the other's benefit" is how spousal top-ups work while both are alive, not how survivorship works.) And the arithmetic gets crueller from there: the year after a first death, the survivor files as single — same house, similar bills, but tax brackets and Medicare thresholds roughly cut in half. In one of my test households, federal tax went from $0 while both spouses were alive to about $2,616 the year after the first death, on less income. Planners call it the widow's penalty. I'd call it the strongest argument for the higher earner delaying their claim — seen properly, that decision isn't a bet on your own longevity at all. It's life insurance for whichever of you lives longest, because the delayed credits survive into the survivor's benefit.

Part 3 models the first-death transition as the fiddly checklist it actually is: joint filing kept for the year of death and single filing from the next year; the year-of-death required IRA distribution still taken; accounts rolled to the survivor exactly once; the step-up in cost basis applied (with a setting for community-property states, where it's full rather than half); pensions cut to their survivor percentage; household spending reduced by a factor you choose rather than one I impose. The survivor benefit calculation includes the awkward corners — the deceased's delayed credits carry over, and there's a statutory 82.5% cap that binds when the deceased had claimed early. Every claiming scenario is tested against the same simulated markets and the same simulated lifespans, so when the comparison table prefers one, the difference comes from the strategy, not from luckier dice.

Answer: b — roughly 48%. And of all older adults, about a quarter will need paid care for more than two years, which means roughly half of those who need care need a lot of it. These are the odds the brochures skate past, and they're the reason Part 4 exists. I built it last, and as its own app — Ballast — because the spec's first rule was that a trustworthy Part 1 beats an impressive but untestable everything-app. Ballast takes the plan you export from the planner and runs the deep analysis: long-term care modelled with age-conditional odds (your risk depends on how old you are, not a coin flipped at 65), costs that progress from home care to assisted living to nursing care and inflate faster than groceries, insurance modelled with its real mechanics of elimination periods and benefit limits rather than a vague "covered" flag.

Ballast also weighs the ways of guaranteeing the essential bucket: a ladder of inflation-protected bonds modelled as actual bonds maturing year by year, or an annuity from a quote you type in yourself, dated, because quotes are perishable goods. And when it searches across all the levers at once — allocation, claiming age, Roth schedule, income floor — it ranks plans by one transparent principle: protect the essentials first, then lifestyle, then legacy, and prefer the simpler plan when results are close. Every component reported separately. If your plan needs more return than your risk capacity can support, it says this plan is infeasible under current inputs rather than hiding the conflict behind a cheerful score. Across all four parts, that's the house rule: no single mysterious number, ever.