fr the landing machine · working name

fintech landing pagesbuilt from your brand book,not from a model's taste.

forj builds and tests the pages inside your perimeter; nothing publishes without passing the gate.

runtime
deploy: your-vpc-eu-1
egress: 0 bytes
models: in-perimeter
brand.json
{
 "accent": "#0A5CFF",
 "radius": 12,
 "tone":   "assured"
}
Open account
Aa — assured
ledger signed
14:02 variant C shipped · hash 9f2e…
14:02 approved — brand owner
13:57 copy diff +12 −4
card contents are demo data
before / after

what forj does to a model's defaults

left: a model with no rules. right: the same hero rebuilt to the brand book of tarn — a fintech invented for this demo — each change signed by the rule that produced it.

specimen_generic.html fails the gate
Revolutionize Your Finances with AI-Powered Insights
Join thousands of users transforming the way they manage money.
Get Started FreeWatch Demo
$2.4Msaved by our users
⚡Lightning Fast
🛡️Bank-Level Security
tarn_brand.html passes the gate
tarn · treasury ops
Move treasury across 14 currencies without leaving a spreadsheet trail.
Tarn settles same-day in 14 corridors, priced on the interbank print, logged to the minute.¹
See the rate ladderTalk to treasury ops
¹ claims-ledger#TARN-CL-014 · Figures are illustrative test datadata-synthetic
tarn — fictional brand
  • purple gradient heroL0 · banned: purple-gradient hero
  • $2.4M with no sourceL3 · truth: unmapped claim → fail
  • hero metric + card gridL0 · banned: hero-metric, card grid
  • paper + navy from tokensL1 · color.tokens, strategy: restrained
  • rate ladder instead of promisesL0 · signature element + L1 atmosphere
  • figures footnoted to the ledgerL3 · claims-ledger#TARN-CL-014

the left pane is a quarantined specimen. tarn is fictional; all figures are demo data.

the alternatives

testing tools split traffic. ai builders live in their cloud. forj does both jobs in your perimeter.

swipe → forj

forj compared with a/b testing tools and ai site builders across five capabilities.
capability testing tools vwo·optimizely ai builders lovable·bolt both jobs · your perimeter
frj
builds landing pageswho produces the page itself no. splits existing pages only yes. yes, in their cloud yes. yes, inside your perimeter
variants as hypotheseswhat a variant is allowed to change partly. traffic split, no hypothesis no. no experiment engine yes. one declared axis, diff-checked
brand as inputwhere the visual rules come from no. out of scope partly. model taste, drifts per run yes. brand.json, every value sourced
runs in your perimeterwhere the code and the data sit partly. cloud, enterprise tiers only no. vendor cloud only yes. your infra, your network
audit trailwhat compliance can read afterwards partly. test logs, no artifacts no. no audit artifacts yes. gate report + claims ledger

positioning, not a competitor audit — public materials, august 2026.

the line · precedence, not decoration

four stations of the press

L0 · craft floor

the floor holds on every variant

contrast, working states, no worn-out fintech patterns.

L1 · brand as input

extracted, never invented

every value cites a source; gaps become proposals.

L2 · one hypothesis

each variant moves one axis

brand rules and disclosures never move.

L3 · the gate

a failed check names rule and fix

no ledger entry, no number on the page.

on conflict:›compliance›brand›craft›variation  — a required disclosure displaces a prettier hero, never the reverse.
one axis moves

when variant b wins, you know why

EXP-042
sample experiment axis: proof metric: application_submit min runtime: 14 days

if the hero leads with an audited savings figure instead of a testimonial, then application submissions rise, because visitors from paid search arrive primed for numbers.

−hero.proof = testimonial · A control
+hero.proof = audited figure #TARN-CL-014 · B
+hero.proof = license + figure #TARN-CL-014 · C
 everything else · 0 diff · brand tokens held constant

a change to anything outside the declared axis fails the gate.

the gate

your approver sees evidence, not a promise

gate_report · EXP-042-proof-C · 14:02 blocked
§1 staticbrand fidelity — 41/41 values resolve to brand.jsonpass
§2 staticcraft floor — contrast 4.9:1 min · 4 widths · 1.2s LCPpass
§3 criticfirst screen — what / why / do, desktop + mobilepass
§4 static+humantruth & compliance — 1 unmapped claimfail
rule L3.claims.mapped — a figure has no entry in the claims ledger
fix add the ledger entry, or cut the figure
blocked until brand owner + compliance sign
§5 diffvs control — only hero.proof moved · 0 other diffpass

demo report · this archive is what an examiner receives

onboarding

the brand is approved before the first landing page, not after

fr

your brand book in. tested pages out. nothing leaves the perimeter.

rulesenforced by the gate
perimeteryours
tarnfictional
figuresdemo data