Skip to content
Start free

ResourcesSample project

One idea, followedfrom research to a running app.

A real project made in LiteSurface, not a mockup: Push Pantry for cooks, a pantry app for home cooks that became the brand Fridgent. Follow it through all six stages and see what each one made, and what it carried into the next.

Start freeFollow the idea

Captured from a LiteSurface workspace on September 26, 2026. Research and scoring ran in fixture mode; the brand, screens, and code used live models. About this sample

  1. 01DiscoverCited evidence
  2. 02DecideA scored concept
  3. 03BrandA brand kit
  4. 04DesignFlows and screens
  5. 05BuildA real repository
  6. 06ShipReady to publish

Sample project · made in LiteSurface

01Discover

Research the space,and keep every source.

The project, Device-Native Opportunity Scan, set one objective: “Find products enabled by modern mobile device capabilities that a small team can prototype rapidly.” Research stored 4 sources and split them into 18 atomic claims, each with a locator back to its passage.

Where it starts

Project objective
Find products enabled by modern mobile device capabilities that a small team can prototype rapidly.
Starting point
What mobile devices can do today
The sample project’s Evidence page: 4 sources and 18 extracted claims, none unavailable or undated. The sources table lists NoteWave pricing, Why I quit every voice memo app, Personal productivity apps market 2026, and On-device inference framework, each extracted, with its claim count, credibility, and age.
Sample project · made in LiteSurface
Models up to 3 billion parameters run on devices released after 2024 with typical latency under 200 milliseconds per request.
On-device inference framework · Example Platform DocsFact · paragraph 1
Privacy concerns were cited by 38 percent of surveyed users as a reason to avoid cloud-based note apps.
Personal productivity apps market 2026 · Example Market ResearchFact · paragraph 2
Every voice memo app I tried uploads recordings to the cloud before transcribing, which I find unacceptable for work conversations.
Why I quit every voice memo app · Example ForumFact · paragraph 1

Three of the 18 claims, as stored.

The sources in this sample are fictional

This project ran in fixture mode, so research read a small made-up corpus on example.test sites instead of the web. Storage, claim extraction, and citations are the same code a live project runs.

Carries forward to Decide

  • 18 claims, each with an id
  • Every score will cite them
Next: Decide

02Decide

Score the idea twice,then decide.

Of the concepts generated from that research, Push Pantry for cooks is the one this sample follows: “Helps home cooks plan meals from what is in the fridge, using push notifications and motion and activity sensing entirely on the device.” Two evaluators scored it on 9 criteria. Code took the medians, checked 6 gates, and put it at 81.6 out of 100: Validate.

From Discover

Evidence coverage
100% of the evidence-based criteria
Citations
Two supporting claims behind most criteria
Defensibility accrual cites
“Every voice memo app I tried uploads recordings to the cloud before transcribing, which I find unacceptable for work conversations.”
The concept Push Pantry for cooks, for home cooks: plan meals from what is in the fridge using push notifications and motion and activity sensing on the device. Its score panel shows 82, Strong, recommendation Validate, medium confidence, 100% evidence coverage, and a 7-day prototype, above the wedge and the value proposition “Never lose a detail, never upload a word.”
Sample project · made in LiteSurface
The concept’s target user and job: home cooks, with a daily pain at intensity 5 of 5 (details are lost or typed up late, and cloud tools feel untrustworthy for private material), the job to be done as when, I want to, and so I can, and the current alternatives: a voice memos app, a cloud transcription service, and a paper notebook.
Sample project · made in LiteSurface
The concept’s score decomposition: synthesized score 82, Strong, recommendation Validate, medium confidence with 13% disagreement, and 100% evidence coverage, then criteria bars for user pain, why now, differentiation, prototype feasibility, and distribution, each with a weight, a median, one marker per evaluator, and two supporting claims.
Sample project · made in LiteSurface

The evaluators here are stand-ins

In fixture mode, deterministic fixtures fill the OpenAI and Anthropic evaluator slots, so the rationales repeat a template. The medians, weights, gates, and recommendation are computed by the real pipeline.
Every number behind this scoreBoth evaluators, all 6 gates, and every cited claim.

Carries forward to Brand

  • The user: home cooks
  • The promise: “Never lose a detail, never upload a word.”
  • A 4-step core loop
  • An implementation handoff (26 sections)
Next: Brand

03Brand

Name it and brand itfrom the research.

Before drawing anything, Brand Studio wrote a strategy from the concept and its evidence. Then it proposed 80 names and ran availability checks on the 20 strongest: Fridgent was confirmed, with Larderly as the runner-up. All 9 required assets are confirmed, plus 7 optional ones.

From Decide

Target user, sharpened
Home cooks who capture pantry and fridge details on the go and don’t want personal household data leaving their phone
First tagline
“Never lose a detail, never upload a word.”
Pain, quoted from the evidence
“Every voice memo app I tried uploads recordings to the cloud before transcribing, which I find unacceptable for work conversations.”
Brand Studio for the sample brand Fridgent with all nine required assets confirmed. The left rail lists the kit by group: name, palette, typography, logo system, icons, imagery, and applications. The canvas shows the confirmed mood board of pantry jars, fresh produce, and slate and oak textures with the brand colors.
Sample project · made in LiteSurface
The Brand Studio name slot: the confirmed name Fridgent and the runner-up Larderly as candidate cards with trademark, domain, and handle checks and easy-to-say and easy-to-spell scores, under a trademark pre-screen disclaimer. The inspector lists fridgent.com, .io, .co, .ai, and .app with availability and prices.
Sample project · made in LiteSurface
The confirmed Fridgent social card: the white logo and a headline on the slate brand color at left, and an open fridge stocked with produce at right.
Sample project · made in LiteSurface
The confirmed Fridgent primary logo: a slate pantry-jar symbol with a leaf and an orange dot beside a serif wordmark, with previews at 16, 32, and 64 pixels and light and dark monochrome versions.
Sample project · made in LiteSurface
The confirmed Fridgent palette sheet, drawn in code: a slate primary, an orange accent, six neutrals, and success, warning, and danger colors, each with its hex value.
Sample project · made in LiteSurface
From the brand strategy, as stored
Category
A privacy-first meal planning app, with the caregiver archetype
Pillar: Trust before convenience
Home cooks have learned to distrust apps that upload personal details before processing them, so this one never does. Cites the forum claim above.
Voice
Warm, practical, and quietly confident — like a trusted kitchen helper who never oversells and never asks for your data.
How the kit was made

Generated as images

GPT Image 2 at draft quality, 6 credits each: symbol, mood board, illustration style, photography direction, pattern, social card.

Derived in code

From the confirmed pieces: typeset name, palette, typography, primary logo, monochrome and inverse logos, clear space, app icon, favicon set, social avatar, splash screen. No credits.

Checked before you pick

Domains, social handles, and a trademark pre-screen for the 20 strongest names. A pre-screen, not legal clearance.

Carries forward to Design

  • The Fridgent logo and palette sheet
  • Type, tokens, and a theme
  • Every screen is designed in it
Next: Design

04Design

Plan the flows,then design the screens.

Mockup Studio planned the app from the concept: 4 flows and 14 screens, each for mobile and desktop, sharing one navigation (Capture, Records, Search, Settings). Each screen is then designed in the brand, with the Fridgent logo and palette sheet attached to every request.

From Decide and Brand

Persona on every flow
Home cooks
The core loop, as flows
Quick capture → Confirm & edit → Records & search
Brand references
Fridgent logo and palette sheet
The Mockup Studio storyboard for Fridgent: an app structure card with the navigation and shared components, then the Getting started and Quick capture flows as rows of screens, each with mobile and desktop frames.
Sample project · made in LiteSurface
The Mockup Studio screen editor: a generated Fridgent mobile screen showing a voice recording with a waveform and an orange stop button, the flow list on the left, and the Create panel on the right with draft and final quality, candidate count, the estimated credit cost, and the brand assets included as references.
Sample project · made in LiteSurface

How the core loop became flows

  1. 1Starts capture with one tapUser
  2. 2Processes input locally into structured itemsSystem
  3. 3Confirms or edits the draftUser
  4. 4Indexes the record for search and remindersSystem
  • Getting started: Welcome, Notifications, Microphone, Motion, Ready
  • Quick capture: Capture, Push capture, Draft
  • Confirm & edit: Edit draft, Add details, Saved
  • Records & search: Records, Search, Record detail

Where this sample stops

One screen is designed and waiting for review (Quick capture, on mobile); none are confirmed yet. Confirmed screens go to the Builder with their specs, so in this sample only the brand made that trip.

Carries forward to Build

  • The storyboard: flows, navigation, components
  • Confirmed screens (none yet)
Next: Build

05Build

Start the real appwith the research inside.

Push Pantry for cooks is a Full-stack monorepo Builder app with four apps: API (Express), Web (Next.js), Admin (Vite + React), Marketing (Next.js). Because it was created from the concept, its first commit already held the handoff. Importing the design package added the brand (54 files).

From every stage before

docs/handoff/
The implementation handoff, from Decide
docs/design/brand/
Logo, tokens, theme, and fonts, from Brand
The agent’s brief
Title, one-liner, target customer, top features, and brand
The Builder for a four-app monorepo, session ready. The agent chat shows a request to use the brand logo with an element attached from the preview, answered by Build with Sonnet 5 for $0.20 with lint and type check passed. The live preview has Web, Admin, Marketing, and API tabs, with the Fridgent marketing site open.
Sample project · made in LiteSurface
The Builder’s Design panel showing the imported Fridgent brand from docs/design with its logo and palette colors and links to Mockup Studio and Brand Studio, beside the live marketing site preview that uses the Fridgent logo.
Sample project · made in LiteSurface
In the repository, from earlier stages
  • docs/handoff/from Decide
    • README.md
    • 00-product/ (5 files)
    • 01-ux/ (4 files)
    • 02-engineering/ (7 files)
    • 04-quality/ (3 files)
    • 05-execution/ (3 files)
    • feature_list.json
  • docs/design/from Brand
    • README.md
    • brand/DESIGN.md
    • brand/tokens.css · tokens.json · theme.css
    • brand/logo/ (11 files)
    • brand/fonts/ (with their licenses)
  • apps/*/public/favicons, app icons, logo

Paths from the app’s version history: 261 files in the first commit, then 54 from the design package.

  1. BuildSonnet 5$0.20

    “update this to our brand logo”

    The marketing header now renders the installed Fridgent logo instead of the placeholder. One file changed; lint and type check passed.

  2. ChatGPT-6 Lunaunder $0.01

    “Which file renders the marketing site header logo, and what size is it?”

    Answered with the file, the line, and the 32 px height. No files changed.

Where this sample stops

One build request so far. The marketing page in the preview still carries the template’s starter copy, with the Fridgent logo in its header; the pantry features themselves have not been built in this sample.

Carries forward to Ship

  • A repository you can read
  • Every change kept as a version
Next: Ship

06Ship

Publish a version,on accounts you own.

Any version in the Builder can leave as a GitHub pull request or a ZIP of the repository, and every publish is recorded as a build with its report and ZIP. Single-app projects can also deploy a preview with its own Postgres database, on plans that include previews.

From Build

The version you reviewed
With its diff and checks
The whole repository
Handoff and design package included
The Builder publish menu, “Publish this version”: GitHub pull request, Download ZIP, and Deploy preview, with the note that every publish is recorded as a build with its report and repository ZIP.
Sample project · made in LiteSurface

Not published yet

We have not published this sample, so there is no live URL to show you. It is a full-stack monorepo, which ships as a pull request or a ZIP today; deploy previews for monorepos are not available yet.

The thread

Five things we followedfrom start to finish.

Nobody retyped anything between stages: each one read what the stage before it stored.

  • The user

    “home cooks”

    1. DecideTarget segment of the concept
    2. BrandSharpened into the positioning’s target user
    3. DesignThe persona on all four flows
    4. BuildThe target customer in the agent’s brief

  • The promise

    “Never lose a detail, never upload a word.”

    1. DecideThe concept’s value proposition
    2. BrandThe first tagline, then the social card headline

  • One piece of evidence

    “Every voice memo app I tried uploads recordings to the cloud before transcribing, which I find unacceptable for work conversations.”

    1. DiscoverA claim from “Why I quit every voice memo app”
    2. DecideCited by defensibility accrual
    3. BrandQuoted as the pain, and cited by a brand pillar

  • The core loop

    1. DecideFour steps, from one tap to a searchable record
    2. DesignThe Quick capture, Confirm & edit, and Records & search flows

  • The logo

    1. BrandConfirmed primary logo, drawn from the symbol
    2. DesignAttached to every screen request
    3. BuildIn docs/design/brand/logo, then in the marketing header

About this sampleand how far it got.

Every screenshot on this page is a capture of one LiteSurface workspace, taken on September 26, 2026. The concept, the brand, the storyboard, and the Builder app are linked records in the same project, not separate demos.

Research and evaluation ran in fixture mode: a small fictional corpus and deterministic stand-ins for the two evaluators, so the numbers are reproducible. The brand strategy, the generated images, the screen design, and the Builder requests used live models; the costs shown are what those requests cost.

We stopped where the table says, and we would rather show that than dress up a finished-looking app.

  • Discover (done)4 sources, 18 claims (fictional corpus)
  • Decide (done)Scored 81.6, Validate (stand-in evaluators)
  • Brand (done)9 of 9 required assets confirmed
  • Design (in progress)Planned: 14 screens; one designed on mobile, none confirmed
  • Build (in progress)Started: handoff and brand imported, one build request
  • Ship (in progress)Not published yet

Your idea next.Same six stages, nothing lost.

Start with a question and your own evidence. LiteSurface scores the concepts, designs the brand and screens from the research, and hands all of it to an agent that builds the app.

Start freeSee the sample evaluation
  • Solo and Team are free during early access
  • Bring your own OpenAI and Anthropic keys