A Retirement Planning Tool Built with AI-Assisted Development
From spreadsheets to production web app in one week—exploring what one person can build with modern development tools.
These tools are educational prototypes. They are not financial advice, they make simplifying assumptions, and they are designed to avoid collecting or storing personal data.
Why I Built It
I'd been using a Google spreadsheet for retirement planning for years—modeling different scenarios, adjusting assumptions, testing early retirement timelines. It worked well enough for me.
Then people started asking if they could use it. Sharing a spreadsheet felt clunky—everyone would need their own copy, formulas could break, and it wasn't mobile-friendly.
A web app made more sense: clean interface, works on any device, can't accidentally break the logic. When AI-assisted development tools matured, I had both a real use case and validation that others might find it useful.
So I rebuilt it.
What It Does
A browser-based retirement planning tool that helps you model:
- Savings growth across Bear, Base, and Bull market scenarios
- Range of Outcomes — 500 simulated market paths showing where your portfolio could realistically land
- Social Security income modeling at your chosen claiming age
- Early retirement feasibility with “what if I wait?” delay scenarios
- How long your savings might last — with a survival probability, not just a single projected line
- Voice and text conversational assistant for hands-free data entry
Honest forecasting for retirement planning decisions — showing the range of what’s possible, not just the average case.
What I Built in Three Phases
Phase 1: Core Calculator
The foundation—a retirement planning tool that models savings growth and drawdown. Built with vanilla JavaScript, Chart.js for visualization, and deployed on Netlify via GitHub. Built with Claude Code.
Timeline: 1 hour from concept to production (traditional 3-person team: ~3 months)
Phase 2: Supporting Tools
Companion calculators that answer follow-up questions: emergency fund sizing, savings benchmarks, withdrawal timelines.
What This Taught: Building ecosystems, not just standalone features. AI implements connected systems well when given clear architecture.
Phase 3: VR Experience
An immersive VR version for Meta Quest headsets—exploring how planning tools work in spatial computing. Built with WebXR API + Three.js with hand tracking and voice control.
What This Proved: Emerging platforms are viable faster than expected. Solo developer + AI = what recently required specialty teams.
Phase 4: Privacy-First Redesign & Advanced Forecasting
The original plan for this phase was different. I’d added Firebase authentication — Google sign-in, cloud sync across devices. It worked. It shipped. Then I removed it.
Here’s why: most people open a retirement planner once, run a scenario, and leave. Requiring an account created friction precisely when the tool should feel immediate and low-stakes. More importantly, retirement data is sensitive. A tool that promises “your data stays on your device” is more trustworthy than one that syncs to a cloud database — even if the sync is secure. Removing a shipped feature takes discipline most product teams never find. The product got simpler, faster, and more honest.
Bear, Base & Bull market scenarios — Rather than asking users to enter a growth rate (a number most people have no intuition for), the planner now offers three named scenarios: Bear (conservative: 5% growth while working, 3% in retirement), Base (moderate: 8% / 5%), and Bull (optimistic: 11% / 7%). Pick the narrative that fits your risk tolerance; the numbers follow.
Range of Outcomes (Monte Carlo simulation) — A single projected line quietly implies more certainty than the math supports. The planner now runs 500 simulated market scenarios using log-normal returns. The chart shows a shaded band (the middle 80% of possible outcomes) and a most-likely path — along with a plain-language “Likelihood your money lasts to 100” percentage. Honest forecasting, plainly explained.
Social Security modeling — Expected monthly benefit and claiming age now factor into the projection, reducing the required portfolio withdrawal each year SS kicks in. For most users the difference is significant.
Timeline: Three to four sessions of iteration — scenario design, simulation architecture, and plain-language UX were each non-trivial to get right.
Phase 5: Conversational Interface — Text & Voice
The form-based interface, however clear, still felt like a spreadsheet. This phase replaced data entry with conversation.
Text & Voice Assistant — A chat-based assistant lets you describe your financial situation in plain language. Ask questions like “What happens if I retire two years earlier?” or “How does switching to the Bull scenario change my outlook?” and get real answers grounded in your actual numbers — not generic responses. Voice input is live as well; the assistant handles spoken ambiguity (recognizing “bass” as Base scenario, “beer market” as Bear) and populates the form in real time.
Timeline: Text interface in one session; voice assistant across two sessions with ongoing iteration on accuracy and narration.
What I Learned
Development Velocity
3-5x compression is real—but only with discipline (version control, testing, deployment, analytics). AI handled implementation while I focused on product decisions. Iteration cycles compressed from days to hours. But I still owned architecture, UX, scope, and quality.
Takeaway: The constraint shifted from "can we build it?" to "what should we build?" Strategic thinking matters more than implementation speed.
Product Thinking as Multiplier
Speed came from knowing what "good" looks like: domain knowledge (what retirement tools need), design judgment (what makes good UX), quality standards (what "production-ready" means), and product sense (how to prioritize features).
Takeaway: Senior talent with AI tools outpaces junior talent without it. Judgment and orchestration matter more than coding speed.
Quality Still Requires Discipline
Implemented an automated test suite covering 180 tests across all calculation modules, GitHub version control, Netlify CI/CD for deployment, and Google Analytics for insights. AI makes shipping fast easy. Shipping fast AND good requires engineering discipline.
Takeaway: Velocity without standards creates technical debt faster. Best teams ship fast with high quality.
Platform Evolution
Everything works across desktop and mobile browsers, no app stores required, standard web technologies, even VR headsets (Quest browser).
Takeaway: Web-first strategies cover more use cases than ever, including emerging form factors.
What This Is (And Isn't)
This is: An educational planning tool, a learning project for AI-assisted development, and evidence of what modern tools enable.
This is not: Financial advice, a guarantee of accuracy, or collecting/storing personal data.
Markets fluctuate, taxes matter, real life is messy. Use this as a conversation starter, not a decision engine.