Gnomon / How it works
Watch Gnomon check a decision
Pick a question a business might ask its AI, change the numbers, and see each step Gnomon takes: the futures it plays out, how many work, what it tells your AI, and the best plan it finds.
Step 1
Your AI asks Gnomon
run_model("hire", { salary_per_year: 65000, days_per_week: 5, full_time_from_month: null, cash_now: 12000 })Step 2
Gnomon sets up the futures
- Sales in each of the 12 months, about 6% either way
- The whole year running 3% better or worse
Step 3
Plays out every one
1,000,000 futures, in 0.312 s
Step 4
Counts what worked
610,101 of 1,000,000 worked out
Step 5
Answers your AI
Not safe enough yet: try the best plan
The question
“Can I afford to hire someone?”
A café with a busy summer: sales of $534,600 a year, costs of $39,000 a month.
- Futures played out
- 1,000,000
- Worked out
- 61.0%
- 610,101 of them
- Time to run
- 0.312 s
- saved run
- Units charged
- 4
- NZ$0.03 per 1,000
How the futures landed: the lowest the bank balance gets
- Worked out
- Did not
Show as a table
| From | To | Worked out | Did not |
|---|---|---|---|
| -$36,134 | -$33,628 | 0 | 14,280 |
| -$33,628 | -$31,121 | 0 | 5,860 |
| -$31,121 | -$28,614 | 0 | 7,800 |
| -$28,614 | -$26,107 | 0 | 10,310 |
| -$26,107 | -$23,600 | 0 | 12,780 |
| -$23,600 | -$21,093 | 0 | 17,140 |
| -$21,093 | -$18,587 | 0 | 20,570 |
| -$18,587 | -$16,080 | 0 | 25,600 |
| -$16,080 | -$13,573 | 0 | 30,040 |
| -$13,573 | -$11,066 | 0 | 35,030 |
| -$11,066 | -$8,559 | 0 | 40,130 |
| -$8,559 | -$6,052 | 0 | 45,190 |
| -$6,052 | -$3,545 | 0 | 48,990 |
| -$3,545 | -$1,039 | 0 | 52,900 |
| -$1,039 | $1,468 | 33,460 | 22,920 |
| $1,468 | $3,975 | 57,190 | 0 |
| $3,975 | $6,482 | 57,900 | 0 |
| $6,482 | $8,989 | 57,690 | 0 |
| $8,989 | $11,496 | 57,330 | 0 |
| $11,496 | $14,002 | 64,250 | 0 |
| $14,002 | $16,509 | 83,400 | 0 |
| $16,509 | $19,016 | 91,010 | 0 |
| $19,016 | $21,523 | 67,870 | 0 |
| $21,523 | $24,030 | 40,360 | 0 |
Cash in the bank, month by month
- Typical future
- 8 in 10 futures land in here
Show as a table
| Month | 1 in 10 below | Typical | 1 in 10 above |
|---|---|---|---|
| Nov | $12,466 | $16,685 | $21,037 |
| Dec | $20,958 | $27,867 | $35,041 |
| Jan | $27,673 | $36,887 | $46,389 |
| Feb | $31,455 | $42,693 | $54,315 |
| Mar | $30,136 | $43,161 | $56,563 |
| Apr | $24,685 | $39,232 | $54,097 |
| May | $16,113 | $32,035 | $48,384 |
| Jun | $5,490 | $22,658 | $40,364 |
| Jul | -$4,177 | $14,390 | $33,461 |
| Aug | -$11,758 | $8,354 | $28,869 |
| Sep | -$16,199 | $5,473 | $27,663 |
| Oct | -$16,537 | $6,888 | $30,900 |
What Gnomon tells your AI
5 days a week ($5,958 a month full time): works out in 610,101 of 1,000,000 futures (61.0%, about 6 in 10).
- The lowest the bank balance gets: usually $4,799; 1 in 10 below -$17,400, 1 in 10 above $19,270.
- Cash at the end of Oct: usually $6,888; 1 in 10 below -$16,537, 1 in 10 above $30,900.
- The main risk: the year as a whole running low (the odds turn at about 0.987).
What Gnomon assumed (5)
- sales_variation = 0.06 (default): How far a month's sales can land from the plan, as a share (0.06 = 6%).
- year_variation = 0.03 (default): How far the whole year can run above or below plan, as a share.
- keep_above = 0 (default): The bank balance that must never be crossed.
- on_costs = 0.1 (default): Employer costs on top of pay (KiwiSaver, ACC, leave), as a share.
- start_month = 1 (default): The month the person starts (1 = next month).
The best plan
Gnomon can also try the choices for you: every option on the same 100,000 futures, then the leading three a million times each, and back comes the plan that is safe enough and does the most good.
This is what your AI gets
When Gnomon is connected, your AI does exactly this behind the scenes: it sends the numbers, Gnomon plays out a million futures in a fraction of a second, and your AI answers you with the count, the main risk and a better plan if there is one.
Each run shows how likely, not a promise. Gnomon names every number it assumed, so you can change it.