VITERBI
Pick a stack
Choose the path that lets a stranger try the idea on Sunday. Not the one that looks impressive in a list of technologies.
No-code, if the idea is a decision, a form, or a flow. Use this if a person needs to answer questions and see a next step. Start with a Figma prototype, a Google Sheet as the database, and one shareable link. Links: https://www.figma.com, https://www.google.com/sheets/about/, https://www.notion.so
A web app, if someone should click through a working page. Use this if the demo is "type this, see that." One repo, one deploy, one path. Links: https://github.com, https://vercel.com/templates/next.js/supabase, https://supabase.com
Data or machine learning, if the result is a number, a rank, or a prediction you can show. Use this if you already have a small dataset and a question. Do not train a giant model. One chart a judge can read is enough. Links: https://colab.research.google.com, https://pandas.pydata.org, https://observablehq.com
Hardware, if the thing has to move, sense, or sit on the table. Use this if you brought parts. Film the behavior in case the wire comes loose before judging. Links: https://www.arduino.cc, https://www.raspberrypi.com, https://wokwi.com
If two paths seem equal, take the one you can explain without notes. Judges score what runs, not what you planned to learn.
You can mix a little. A Figma screen in front of a real button is fine. A hardware sensor that writes one row to a sheet is fine. What is not fine is four half-builds. Pick one path before dinner on Friday and stay on it unless the problem drop makes it impossible. If you switch, switch before Saturday noon, not at midnight.
Partner tools are optional, not a requirement. IBM Bob is free for 30 days if you want a coding assistant. Terminal49 has live container data if your track is ports. Use them when they shorten the demo. Ignore them when they do not. The stack is a means. The three-minute path is the point.