| Current coursedo nothing · baseline | Bridge · 24%SEPP + fill bracket | Convert 32%derived from Bridge | Trust '36derived from Convert | |
|---|---|---|---|---|
| Goals | ||||
| Income floor | ||||
| College | ||||
| Dynastic trust | ||||
| Metrics · deltas vs. baseline · real 2026 $ | ||||
| Lifetime federal tax | $2.41M | −$190K | −$262K | −$262K |
| To heirs, after their tax | $6.69M | +$212K | +$341K | +$355K |
| Tight-years liquidity | $0.9M | −$140K | −$310K | −$310K |
| Holds under 1966 start | no · 2061 | no · 2066 | yes | yes |
{lede}
: null}{eyebrow}
: null}{lede}
: null}Product
A single projected number is false precision that invites false confidence. The unit of value in Talunavo is the difference between two plans under identical assumptions, and then, once you have chosen, whether reality is still matching the plan you chose.
Plans and scenarios
A plan declares a base and overrides it. The library prints the lineage: Current Course, then Bridge · 24%, then Convert 32% derived from the bridge, then Trust '36 derived from that. Open the third one and its diff against its base is one line: the bracket.
Scenarios are separate from plans. Returns per asset class with per-account overrides, inflation, longevity, contingency resolutions, and the tax regime are editable field by field, each with its plausible band. Saving writes a new version rather than overwriting one, so every projection ever computed stays reproducible from its own inputs.
A validation list under the plan editor reads your goals as well as your plan. "This plan's conversion policy has no tax source while the college earmark holds the taxable account" is a finding neither the plan nor the goal is wrong about.
The ledger
A projection you cannot audit is astrology with a chart library. So every year is projected into a ledger you can read, drill into, reconcile, and export. Every figure drills into the flows and the cited tax lines that produce it, and every flow carries the module that originated it.
Tax law is data: versioned per year and jurisdiction, every figure carrying its citation, its indexing rule, and its confidence tier, content-hashed into the run's fingerprint. Golden files pin every cliff against official publications.
Upload your own spreadsheet's balance-by-year and the app keeps it, re-running the diff on every review and flagging divergence beyond a tolerance, stating exactly which lines both sides model.
With room. The margin tells you how much, in the goal's own units: dollars per year for a floor, a lump sum for a legacy.
Funded, with little cushion. The margin says how little, and the binding year says when the plan is tightest.
Below the funded test. The gap is paired with its effect and its cheapest lever. This is the only state that wears the decision red.
The app cannot judge this yet. An honest gap, not a warning, and it names the answer that would close it.
The projection engine never knows a goal exists. Evaluation is a pass over ledgers, and everything one ledger cannot answer, such as which goals survive a bad decade or whether a six-year care event is absorbable, is answered by orchestrating more deterministic runs inside a budget of at most 30 runs and half a second. A status chip reads the range across your named scenarios, so "funded under Expected, short under 5% tuition inflation" is one honest mark rather than a flattering one.
Forced distributions, and the first year the forced draw exceeds the spending need. The wall is the thing most households half-know is coming and have never seen drawn.
Re-decided every year against that year's measured tax, with IRMAA, ACA-cliff and 0%-gains guards that trim the amount and say what the guard cost.
All three methods, on an account carved out of the IRA, with the rate cap from data, a duration lock computed rather than typed, and a modification inside the lock modelled as the retroactive 10% plus interest. The save refuses it rather than warning about it.
With spousal and survivor benefits, and a cost-of-living adjustment that the scenario's own inflation drives.
What each strategy does not model is written down in the product rather than left to be discovered.
The Explorer
Every other screen answers "is this goal funded?" The Explorer answers "what is fundable?" as a grid of candidate worlds that streams in as each cell is computed. One cell is one projection and one evaluation through the same evaluator the goal board uses.
The grid asks your household's two questions, not two fixed ones. Still earning: retirement year across the top, income floor down the side. Already retired: spending down the side, longevity across. A grid with fixed axes would ask a 68-year-old widow when she plans to retire.
Tap a cell for the sandbox behind it. Adopt it and the goal is born carrying the cell it came from, so a goal reached by exploring can always say what question it was the answer to.
Checkpoints and watches
Adopting a plan freezes a baseline: the resolved plan, the assumptions, the balance sheet, the run's identity, and one line you write about why. It cannot be edited, only superseded, and the chain of superseded baselines is a decision journal nobody had to remember to keep.
Once a year, quarterly at most, a checkpoint measures the year against it. Nine dimensions, each pre-filled with the planned figure so confirming an on-plan year is one click. Missing numbers never block the verdict. The drift is decomposed into spending, contributions, income, tax and returns, and the pieces sum exactly to the total.
The verdict is one of three: on track, drifted but the conclusion is unchanged, or review warranted, the last unconstructible without a named reason. A watch is a decision left running against the real world. When its condition binds it produces a finding, never an execution.
Uncertainty
Deterministic is the default and what plans are compared in. Historical sequences are built to answer "how would this plan have fared starting in 1966?" and every band keeps the year that produced it. Monte Carlo returns a distribution: a percentile fan with the deterministic run pinned through it, the shortfall's magnitude and timing, and the years of spending covered at the 10th percentile.
Progress is reported in trials, not percentages. Click a band edge and the trial behind it is redrawn from its seed, because no trial ledger is stored anywhere, and the payload says so.
Home becomes a solvency verdict instead of a board. Does the money hold, in which of your named scenarios, and where does it stop? Sustainability tiles are drawn from the same rules layer, each citing your own numbers.
Estate routing, an exemption ledger, trusts, and a gift modelled as the two flows it really is, at transfer-date valuation and carryover basis. Every figure these touch is labelled hypothetical with its assumption cited.