Inspiration

A $600 purchase can look affordable today and leave you short when rent arrives next week. A credit card can keep cash in your account while adding costs you don’t fully understand. Investing can feel intimidating before you even make your first contribution. We built Forespend to make those consequences visible. We wanted financial education to become something people can explore, question, and apply to their own decisions.

What it does

Forespend combines a financial calendar, interactive learning games, investment simulations, and a voice-enabled AI assistant. Users can:

See their money over time: view income, expenses, and daily balances on a calendar or accessible list.
Test a purchase: compare cash, installments, and credit to see payment dates, extra costs, remaining debt, and potential cash shortfalls.
Choose a cash cushion: see which options fall below an amount they want to keep in checking.
Learn by experimenting: complete credit and investing quests with predictions, simulations, and knowledge checks.
Explore investing: model recurring contributions, compare hypothetical fund fees, and examine adverse-return scenarios.
Ask questions by text or voice: receive explanations grounded in the current simulation and listen through ElevenLabs. Our prototype uses a fictional customer’s Capital One Nessie sandbox accounts, alongside clearly identified sample transactions and future estimates. Simulations do not move money or execute investments.

How we built it

We built the frontend with React, TypeScript, and Vite, using a shared visual style across the calendar, learning games, and investing tools. Our FastAPI backend connects to Capital One’s Nessie API and runs the financial simulations. Python handles the calculations using decimal arithmetic, keeping cash balances, debt, and investment contributions separate. Anthropic’s Claude API explains the calculated results and answers follow-up questions. The assistant receives relevant scenario information rather than relying on financial totals supplied by the browser. ElevenLabs powers speech-to-text and text-to-speech. Users can record a question, review its transcript before sending, and listen to the displayed answer. We reviewed the help of Claude and Chatgpt, iterated on the experience, and checked calculations against explicit examples.

Challenges we ran into

One challenge was making financial comparisons fair. Credit can leave more cash available initially, but that does not mean the purchase costs less. We needed to show both available cash and future obligations. We also had to avoid double-counting transactions, handle installment rounding, distinguish accrued interest from future interest, and keep historical activity separate from projections. The interface took several iterations. Our first version displayed plenty of information but made the useful actions difficult to discover. We simplified the calendar, added explanations beside unfamiliar terms, and moved voice into a shared assistant drawer. Integration brought practical challenges too: an old backend process returned an outdated response format, a Windows environment variable overrode our intended API key, and our first ElevenLabs voice required a different plan. Diagnosing these taught us to check the running system as carefully as the code.

Accomplishments that we're proud of

We connected learning to action: a credit lesson can open a purchase comparison, and an investment experiment can place contributions on the calendar. We also built:

A calculation engine checked through automated tests and live API verification.
Consistent payment comparisons that show costs and obligations alongside cash.
Fund-fee illustrations that separate contributions, deducted fees, and lost growth.
Real voice transcription and playback.
Keyboard-accessible controls, contextual term explanations, and text alternatives to visual information.
An assistant that marks answers outdated when the underlying scenario changes.

What we learned

Financial clarity requires more than correct arithmetic. Dates, wording, assumptions, and presentation all affect whether someone understands a result. We learned to keep calculations in deterministic code and use AI to explain them. We also learned that accessibility improves the whole experience: clearer labels, readable comparisons, keyboard support, and voice alternatives help people with different needs and levels of financial knowledge. Most importantly, we learned how to turn separate features into one connected journey: learn a concept, test a decision, see its consequences, and ask why.

What's next for Forespend

We want to expand recurring income and expense planning so users can explore longer periods with more complete budgets. We also plan to add persistent saved plans, broader financial scenarios, and further accessibility testing with users and screen readers.

Beyond the hackathon, we envision Forespend as a financial learning and planning experience for students and first-time earners, helping them understand everyday credit decisions and approach investing with clearer expectations.