FoodExample project · Fictional nonprofit
How does more food arrive on time with the same volunteer hours?
An organization connects surplus food from shops with receiving organizations. Every afternoon someone has to decide who picks up what and where it goes before it expires or the center closes. The first goal is to help prepare a workable pickup plan that a coordinator reviews and adjusts.
- Nonprofit
- Food recovery network
- Led by
- [TBD]
- Repository
- Opens with the project
Six pickups about to expire, three volunteers, three centers. Who goes where?
The problem.
The problem, in the nonprofit's words"Food offers all arrive at once, mid-afternoon, and I have two hours of volunteers and twelve centers with different hours. If a pickup fails, that food is lost."
1,125 million kg
of food wasted in Spanish households in 2024
even after a 4.4% drop from the year before
MAPA · “El desperdicio alimentario en España se reduce el 4,4 % en 2024” (Aug 1, 2025)
8.04 million kg
of fruit and vegetables distributed by a single food bank, Madrid's, in 2024
47.5% of everything it distributed: almost half, perishable
Fundación Banco de Alimentos de Madrid · “Datos básicos cierre 2024”
Friday, 6:30 p.m. 34 surplus notices have come in, and the bread expires tonight. There are 9 volunteers available, and two of the receiving centers close at 9 p.m. At 7:15, two volunteers from the same area cancel. The coordinator reassigns what she can by phone and, by the end of the evening, several pickups never happen.
Pick a task.
Three tasks, each with what you learn. Open one to see the steps and requirements.
- 01
DefineM
Map pickup and delivery constraints
You learn: How to turn what someone tells you into rules a machine can use, with AI to organize them.
Know what constrains each pickup and delivery.
What you do
- 1.Talk to the coordinator and two centers.
- 2.List hours, expiry, cold chain, capacity and zones.
- 3.Separate what's never broken from what's negotiable.
Accepted when
- The coordinator confirms nothing important is missing.
- Each constraint says whether it's hard or negotiable.
- Three real afternoons described, no personal data.
- You deliver
- A document with the constraints classified and examples.
- What credit it gives
- Your reviewed contribution record. If it's your first on this project, its badge too.Technique you show: Defining with AI: from conversation to constraints.
- 02
BuildL
Build the pickup planner
You learn: How to orchestrate several agents in parallel (one builds, one tests, one reviews) without blowing the budget.
Propose a feasible pickup plan and redo it when someone cancels.
What you do
- 1.Model the constraints as data.
- 2.Propose a plan that respects the hard constraints.
- 3.Redo it after a cancellation, explaining each change.
Accepted when
- Never breaks a hard constraint.
- Collects more food than the simple rule, and says how much.
- Each assignment comes with its explanation.
- You deliver
- Code with tests and a short results report.
- What credit it gives
- Your reviewed contribution record. If it's your first on this project, its badge too.Technique you show: Orchestrating agents in parallel with tests.
- 03
TestM
Test cancellation scenarios
You learn: How to design edge cases with AI and measure, repeatably, what a system does.
Check that the plan holds up on hard afternoons.
What you do
- 1.Build eight scenarios from real afternoons.
- 2.Run the planner on each one.
- 3.Describe where it fails or surprises.
Accepted when
- Includes chained cancellations.
- Runs again with a single command.
- Each failure comes with the scenario that causes it.
- You deliver
- Reusable scenarios, a script and a report.
- What credit it gives
- Your reviewed contribution record. If it's your first on this project, its badge too.Technique you show: Testing with AI: repeatable scenarios.
When it's done.
A pickup plan proposed in minutes that respects schedules, expiry times and each center's capacity, which the coordinator reviews and redoes when someone cancels. The organization would measure it: food collected and time to prepare the plan.
- Bring shop notices, volunteer availability and center opening hours into one place.
- Propose a pickup plan that the coordinator reviews and confirms before sending.
- Redo the plan when someone cancels, without losing what's already confirmed.
- Explain each suggestion: why this person and this center.
Kilos collected versus reported, and minutes to prepare each afternoon's plan, before and at 90 days.
Risks and sensitive data
- That it proposes impossible routes: each proposal is explained in one sentence and the coordinator decides.
- That it depends on a paid service the organization can't maintain.
- No data about the people the receiving centers serve. About volunteers, only availability and area; no personal addresses.
Data and license
Apache-2.0. Shops, centers and routes come from open data (OpenStreetMap, OSRM); test scenarios are invented and say so.