87ns
Per-record sensitivity propagation: 4-arg typed DAG.
bench · commit 46a6328 →Financial models
as code.
scroll ↓
live · system pulse0 compiles · 0 in last window
4,000drivers
Largest validated DAG. Compiled cold in under one second.
bench · scale-warm →100/ 100
Institutional Excel-export rubric. Three-pass compile, no drift.
rubric · six-sheet test →bit-identical
Same IR always produces the same outputs. Deterministic, recomputable.
determinism test →MCP+ REST
Both protocols, day one. Your agent calls the same engine your CFO can curl.
api reference →50/ month free
Then $0.10 per answer. No seat licenses. No proposal.
pricing detail →Not a spreadsheet.
Not a chatbot.
A typed compilation layer.
Spaceland · consumer hardware + subscription · fictitious demo model
compiling the model…
Every answer
comes with its receipts.
Specimen · what your agent hands back
seed_saas_v3.xlsx · Flatland export100 / 100 · rubric-pass · 3-pass · bit-identical
D9|
=D7-D8 [ EBITDA = Gross_profit − Opex ]Q1 2026
Q2 2026
Q3 2026
Q4 2026
Revenue
$ 96,000
$ 102,720
$ 109,910
$ 117,604
MRR
$ 8,000
$ 8,560
$ 9,159
$ 9,800
Net new logos
12
13
13
14
Churn
6.0%
6.0%
6.0%
6.0%
COGS
$ 24,960
$ 26,707
$ 28,577
$ 30,577
Gross profit
$ 71,040
$ 76,013
$ 81,333
$ 87,027
Gross margin
74.0%
74.0%
74.0%
74.0%
Opex
$ 62,400
$ 65,520
$ 68,796
$ 72,236
EBITDA
$ 8,640
$ 10,493
$ 12,537
$ 14,791
Cash runway (mo)
14.2
16.1
18.4
21.3
✓ 5 / 5 assertions passing·12 drivers·3-pass compile·typed: Currency, Percentage, Count, Duration, Ratio
Cover
Inputs
Model
Assertions
Scenarios
Checks
Live formulas via named ranges. Blue inputs, black formulas. Six sheets, no drift between them. The same artifact your auditor receives, generated from a typed compilation graph; not assembled by hand.
Why we still export to Excel →Four ways inpick the surface that fits the work
01 · Founders→
Your Series A model, compiled.
Describe your business in natural language. Out comes a typed model with named assumptions, named ranges, and an Excel your investor will read. Thirty minutes, not three weeks.
≤30 min
first compile to investor-ready Excel
02 · Finance Engineers→
The compile layer the category runs on.
Compiled answers across your engagement roster. Typed drivers, recomputable models, audit-trailed exports. Built for the firm that bills as fast as it can defend the work.
100 / 100
institutional Excel rubric, every export
03 · AI-Native developers→
A financial modeling MCP your agent can call.
Typed IR. DAG. Sensitivity. Scenarios. Excel. One npx command and your agent harness has a deterministic finance backend. No custom-tool glue, no JSON-pretending-to-be-a-model.
MCP + REST
both protocols, day one
04 · Platforms→
The typed substrate platforms embed.
If you're building a finance product on top of LLMs, Flatland is the deterministic layer underneath. We sit below your agent and behind your UI: IR in, audit-grade artifact out. You keep the customer; we hold the math.
OEM-able
embed the engine, brand the surface
Quantitative reasoning,
compiled.
Flatland · index of everything
Comparisons
Disasters
© 2026 Flatland · made for systems of record · live pulse · awaiting first compile