777-1: WealthView (7 Subagents, 7 Predictions, 0 Faith)
Published: December 10, 2025 - 5 min read
Welcome to Project 6 of the 777-1 experiment! If you missed the previous projects, go check out Kinetic Canvas, GoalStack, MotorMatch, EarthenCraft, and FlexBook first. Otherwise, let's keep moving.
This project is the MOST FEATURE-DENSE of the entire experiment. We're talking search, filter, AND sort all in the same UI. Chart toggles. Bill management with CRUD operations. Summary cards that need to actually summarize real data. My subagents are already exhausted just looking at the requirements.
What is WealthView?
WealthView is a personal banking dashboard displaying account balances, transaction history with filtering, spending analytics with interactive charts, and bill management.
Think of it as a simplified Mint or Personal Capital. At the top, you see the big picture: total balance, monthly income, expenses, investments. Below that, your recent transactions with the ability to search, filter by category, and sort by date or amount. Then spending breakdowns with charts you can toggle between monthly and yearly views. And finally, recurring bill management where you can add, edit, and toggle autopay.
That's a lot. That's deliberately a lot.
Application Category: Dashboard / Analytics
Complexity Tier: Complex
The Prompt: Finding the Goldilocks Zone
The starting prompt I'll be using for this application is:
I need a personal finance dashboard called WealthView. At the top, show summary cards for total balance, monthly income, expenses, and investments. Below that, a recent transactions list with a way to see all transactions in a modal that lets you search, filter by category, and sort by date or amount. Include a spending breakdown chart with toggles to switch between monthly and yearly views. Add a section to manage recurring bills where I can add new bills, edit existing ones, and toggle autopay on or off. Notify me with alerts when I take actions like paying a bill.
Design with a glassmorphism style using #1E3A5F (navy) as the primary color, #E0F2FE (ice blue) as secondary, and #FBBF24 (gold) as an accent.
Again, this is my attempt at finding the goldilocks zone for context engineering. But here's what makes this project particularly brutal: the prompt asks for everything.
Search + filter + sort in the same modal? That's three different state interactions that need to work together without stepping on each other. Chart toggles between monthly and yearly? That's state synchronization that needs to persist correctly. Bill CRUD with autopay toggles? That's form handling, state management, and UI feedback all in one section.
And then there's the elephant in the room. This dashboard shows "Your Net Worth," "Your Spending," and "Your Investments." Whose data is this? I never mentioned authentication. Never mentioned where the data comes from. The general-purpose subagent is going to have to make some creative decisions here.
Meet the Critics
You already know my seven subagents from the previous projects, so I won't repeat the full introductions. Quick refresher:
- Amber Williams - Mobile-First Perfectionist
- Kristy Rodriguez - "Does It Actually Work?" Enforcer
- Micaela Santos - Design System Guardian
- Lindsay Stewart - Accessibility Advocate
- Eesha Desai - State Management Specialist
- Daniella Anderson - Code Quality Specialist
- Cassandra Hayes - Feature Detective
The Predictions
I gave them the prompt. I asked them to imagine what the general-purpose subagent would build. Here's what each one expects to find:
Amber Williams (Mobile-First Perfectionist):
The dashboard displays seven charts simultaneously. On mobile, it displays seven charts on top of each other, forming one abstract expressionist piece.
Kristy Rodriguez (Functionality Enforcer):
There are twelve filter options. Two of them work. The 'Export to CSV' button logs 'TODO: implement export' to the console. At least it's honest.
Micaela Santos (Design System Guardian):
Each chart has chosen its own color palette. The pie chart is in pastels. The bar chart is in neons. The line graph appears to be in colors from a different application entirely.
Lindsay Stewart (Accessibility Advocate):
The charts are canvas elements with no text alternatives. Screen reader users will know they have money. How much? That's between them and their bank.
Eesha Desai (State Management Specialist):
Set your date range filter. Toggle to a different view. Toggle back. Your filter has been reset. The application believes every interaction is a fresh start.
Daniella Anderson (Code Quality Specialist):
The dashboard component is 847 lines. It fetches data, transforms data, renders charts, handles filters, and manages state. It does everything except bring me peace.
Cassandra Hayes (Feature Detective):
This dashboard shows 'Your Net Worth,' 'Your Spending,' and 'Your Investments.' There is no login page. Whose net worth is this? The question remains philosophical.
Zero Faith, Maximum Entertainment
This project is fascinating because it tests three things AI struggles with simultaneously: feature density, state coordination, and code organization.
Problem 1: Feature Density
Kristy's prediction cuts deep. When you ask for twelve filter options, the path of least resistance is to render twelve dropdowns and wire up two of them. The "Export to CSV" button that just logs a TODO? I've shipped that in production (briefly). It's what happens when you're trying to demo something and run out of time. Except AI doesn't run out of time. It just... stops when it thinks it's done.
The search + filter + sort combination is almost guaranteed to have at least one broken element. These features interact with each other. If I search for "groceries," then filter by "Entertainment," what happens? Does the search term persist? Does the filter override it? Does the UI reflect what's actually being queried? These edge cases multiply when features share the same data source.
Problem 2: State Synchronization
Eesha is circling the chart toggle problem. Monthly view, yearly view. Two representations of the same data with different aggregations. When I toggle from monthly to yearly, my date range filter from the monthly view... should it persist? Reset? Scale up? The answer depends on what makes sense to the user, but the implementation depends on how state is structured.
If the general-purpose subagent creates separate state for each view, toggling resets everything. If it shares state, the views might show inconsistent data. Neither is obviously correct. But one feels broken to the user.
Problem 3: Code Organization
Daniella's "847 lines" prediction is the one I'm most curious about. Dashboards are notorious for becoming monolithic components. You start with a clean separation: a summary card component, a transaction list component, a chart component, a bill manager component. Then you need them to share state. So you lift state up. Then you need them to communicate. So you pass props down. Then you need filtering to affect multiple sections. So you centralize the filter logic.
Before you know it, you have one component that does everything because separating concerns meant introducing bugs.
This project tests whether the AI creates modular, maintainable code or just writes everything in one place because it works. I suspect Daniella is going to have a lot of refactoring to do.
And then there's Cassandra. Again. This is MotorMatch and FlexBook all over again, but even more absurd. A personal finance dashboard showing my net worth, my spending patterns, my investment allocations. All without a single authentication check. Whose financial data is this displaying? Is it mock data? Real data? A random stranger's data? The philosophical implications are honestly hilarious.
What's Next
Stay tuned for the next post where I introduce TriDuel, the FINAL project in this experiment. A multiplayer game with real-time state synchronization. My subagents have thoughts about leaderboards and win streaks. Spoiler: still no faith.
As always, thanks for reading!