DESIGN-GOAL FIT TRIBUNAL  ·  THE SCREEN v. THE GOAL CASE № DGF-001  ·  SESSION: OPEN
Design-Goal Fit
Every screen,
on trial.

Design tools stop at “does it look good.” Fustat puts a screen on trial against the goal it’s meant to move. It stamps a verdict, and shows its evidence.

Put a screen on trial →open-source · no signup to try
↑ Goal: checkout conversion
＋ ～ ＋
① The verdict · live
The charge · goal this screen must serve
PARTIAL · 3/5
90%CERTIFIED
Critique confidence. A second, read-only agent graded this verdict. The reviewer never grades its own work.

Pick a charge, and the court re-decides. That’s the plugin.

§ · Fustat at work

The owls never clock out.

Point Fustat at your screens and a farm of reviewers goes to work. Every screen gets a verdict, every verdict gets evidence, and the critique never sleeps.

Exhibit A · the verdictFustat impact review: Design-Goal Fit and Critique Confidence scores
Exhibit B · the flowFustat flow with a rendered screen
Exhibit C · the hubFustat hub with projects
FIX 9Collapse the mood picker
?How did I get this?
EVALThe reviewer never grades its own work
§1 · Critique · the plugin

Sketch → screen → verdict.

i · bring

Any design, any fidelity

A napkin, a screenshot, a Figma frame. Fustat learns your design system and renders it in your look.

ii · judge

Against the goal, not taste

Name the KR. Every choice is weighed against it, with “how did I get this?” one click down.

iii · vouch

A second agent grades it

A read-only judge certifies the critique’s quality. A confidence seal you can take to leadership.

§2 · The hearing

Design says one thing.
The business says another.

Plaintiff
Design
↑ THE GOAL
presiding
Defendant
Business
POW

They can argue all day. Fustat ends it. Every screen judged against the goal, not the loudest opinion in the room.

§3 · Fustat Cloud · the hosted version

Don’t just judge it. Put your users on it.

Give Fustat a persona and a cohort size. It deputizes a thousand agents for one, four hundred for another, and walks every one through the screen. A sequestered holdout never sees the change, so your lift has a clean control, not a hunch.

The jury pool · persona cohorts1,650 agents
Seat your cohorts
M
Mona · price-sensitive
compares before she commits
×1,000
O
Omar · in a hurry
bounces at first friction
×400
D
Dina · skeptical
needs a reason to trust
×250
⊘
Holdout · 10% sequestered. This control cohort never sees the new screen. Fustat Cloud carves it off and splits the traffic. You never wire it up.
Testimonyidle

Run the cohorts to see each one walk the screen, at scale.

Converted · treatment vs holdout
saw v2
·
holdout
·

The holdout stayed on v1 and converted at the old rate. So the gap is the change, not the season. Directional rehearsal, not a guaranteed lift; Fustat tells you which calls are big enough to confirm with real users.

§4 · Human in the loop

Your experts aren’t
overruled by an agent.

A senior designer disagrees? They file a note on the verdict: [accepted] or [needs work]. Fustat reads it on the next hearing and amends its finding. The expert leads; the agent learns.

THE HUMAN STEERS. FUSTAT ADAPTS.

Expert note · rec 2
S
Sara · Design Lead
“Agree the streak is brittle, but don’t add a freeze. We tested it, it kills urgency. A grace message is better.”  [ACCEPTED]
Fustat
Understood. Amending rec 2, so the next hearing recommends a grace message tied to the streak instead.
§5 · Why it’s believable

One AI writes it. Another evals it.

Would you take AI feedback to your VP? Only if something vouched for it. One agent writes the critique; a second, read-only agent evals it, opening the real screen to verify every claim, then stamps a confidence score you can defend.

The evidence
Your screen
KR: Day-7 retention
Agent 1
The critic
writes the critique
against the goal
The critique
Findings + fixes
ranked by impact
Agent 2 · read-only
The judge
evals the critique,
not the design
Confidence
26/30
PASS · 87%
6 dimensions,
a bar you set
the judge opens the real screen to verify every claim
READ
ONLYno edits
The reviewer never grades its own work.
The rubric

The eval is yours to set.

The bar is yours. The rubric is a file you own. Edit any dimension, set its weight, add your own. Try it →

THE EVAL ISN’T A BLACK BOX. IT’S YOURS TO SET.

Your eval · critique rubric
27/30
PASS90%
pass mark · 80%
§6 · Choose your counsel

From “run it yourself”
to “let Fustat handle it.”

Afraid of handing design to an agent? Don’t be. Fustat starts fully open and fully in your control. Run it yourself, inspect every line, keep your data on your machine. You step up a tier only when you want us to do the heavy lifting.
Tier 01

Open source

Run it yourself. You hold every key, inspect every line.
Free
self-hosted · forever
  • + The full critique + eval method
  • + Your machine, your compute
  • + Apache-2.0
  • ✕ No cloud agents
Run it yourself ↗
Tier 02

Free

Try it hosted. No setup, no card.
$0
3 critiques to start
  • + 3 full critiques, free
  • + Design system · render · score
  • + Nothing to install
Start free
Tier 03

Starter

For PMs putting a screen on trial every week.
$5
per month · 10 critiques
  • + 10 critiques every month
  • + Every project saved + reopenable
  • + Top up anytime
Get Starter
Coming soon
Tier 04

Pro

Persona agents, A/B testing, Figma in, team hub.
$$
per seat / month
  • + Persona agents + testing
  • + Figma in · team hub
  • + Everything managed
Coming soon
Put a screen on trial

Critique it. Test it.
Trust it.

Start with the open-source plugin. Get first access to Fustat Cloud: Figma in, agents that test your design, zero setup.