Conditional

Open source · Append-only ledger · It grades its own forecasts

Every money question you have is an if.

If I keep spending like this. If the market does that. If I take the job. Conditional reads what you actually spend, finds what you are locked into, and simulates the answer a thousand times over — then checks its own work against what really happened.

Not stock imagery. Each panel is generated from a real series above — geometry from its volatility, tilt from its trend. Same data always makes the same picture.

Stays above $500 in 100% of simulated futures

1,200 paths · 90 days · 14 obligations found · 658 rows

$0$6.0k$12k$18k$500 floorRECORDEDPROJECTEDMay 27Jul 15todaySep 24
Hover the chart for any day. half of futures nine in ten
In checking

$3,150

available right now

Net position

$13,686

all accounts, card included

90-day median

$16,983

$16k to $18k in 9 of 10

Shortfall risk

0%

never, in any path

Live output from a generated sample household, through the same code paths a real ledger runs.

Why a single trend line is always wrong

01

Committed

−$2,694 / month

Same amount, same day. Known in advance, so the forecast schedules it rather than guessing.

02

Metered

−$196 / month

Arrives on schedule but the amount moves with the season. Forecast from your own history for that month.

03

Chosen

$2,143 / month

No schedule and no fixed amount. This is the part that gets simulated, and the only part you can move.

Regress one line through all three and you get a number describing none of them. Rent does not drift. A January heating bill is not a January-shaped July bill, it is roughly double. Groceries have no schedule at all but do have a weekday rhythm and a long right tail. Model each the way it behaves and the forecast starts holding up.

Found without being told

Nobody entered these. The detector derived them from the ledger.

See the full schedule
MerchantRhythmScheduleAmountPer monthNext
Cedar Ridge Propertyevery 30d · 8×83% reliablefixed−$1,811Jul 30
Northbank Auto Loanevery 31d · 8×82% reliablefixed−$382Aug 14
Statewide Mutualevery 31d · 8×74% reliablefixed−$210Aug 20
Lakeline Energyevery 30d · 8×83% reliablevaries ±24%−$135Aug 20
Relay Mobileevery 30d · 8×76% reliablefixed−$95Aug 12
Fiberlineevery 31d · 8×82% reliablefixed−$89Aug 7
City Water Worksevery 30d · 8×83% reliablevaries ±14%−$62Aug 6
Ironworks Gymevery 31d · 8×82% reliablefixed−$43Aug 4

The conditional mood, computed

You supply the if. The engine propagates it through obligations you actually have.

Run them yourself

Inflation picks back up

If prices start climbing at 6% again

90-day median
−$383
Shortfall risk
+0%
Runway
unchanged

A hard winter

If heating costs run 35% above normal

90-day median
−$403
Shortfall risk
+0%
Runway
unchanged

Income stops

If I lost my job tomorrow

90-day median
−$16,505
Shortfall risk
+48%
Runway
unchanged

The outside world, live

Pulled from public APIs at request time, not a cached screenshot.

All 18 series

S&P 500

Stooq

Relevant to a household only through what it does to a position you actually hold. Here it feeds the scenario engine, not a ticker widget.

No data returned

Stooq returned no usable observations

Stooq

USD to EUR

European Central Bank via Frankfurter

0.88

+3.3% y/y

For anyone earning in one currency and spending in another, this is a contractual line item, not a curiosity.

0.860.88Mar 18May 25Jul 24

Treasury average interest rate

U.S. Department of the Treasury

3.71%

+12.2% y/y

What risk-free money earns. Sets the floor for whether your savings are keeping up or quietly losing ground.

0.0%3.0%6.0%Jan 31Apr 30Jun 30

CPI, all items

U.S. Bureau of Labor Statistics via FRED

The general price level. Rises here scale the whole discretionary block, because last year's grocery days no longer cost what they cost.

No data returned

FRED (St. Louis Fed) needs FRED_API_KEY in your environment. It is free to obtain — see the sources screen for the link.

U.S. Bureau of Labor Statistics via FRED

6 of 18 catalogued series are ready in this deployment. 6 providers need no key at all, which is why a fresh clone has usable market data immediately.

Proof, not assertion

It grades its own forecasts, and it has failed twice in public.

Walk-forward validation ships as a screen, not a footnote. The model is placed at points in the past, handed only what was knowable then, and scored against what actually happened. It has already caught a synthetic-data defect and a systematic optimistic lean traced to credit card payments being misclassified as internal transfers. Both are written up in the roadmap rather than quietly patched.

See the scorecard
HorizonBand heldClaimsWidthMedian off by
14 days0%90%7%$1.1k
30 days50%90%11%$564
60 days75%90%16%$886
90 days75%90%20%$1.4k

Coverage above 90% means the bands are hedging. Below means they are overconfident. Narrow is only a virtue while coverage holds.

Built for specific situations

A tool that claims to help everyone helps nobody in particular.

All 7

Everything above is live. Go break it.

Eight months of a generated household. Move the spend control and watch the odds move, or drop in your own CSV export.