Inspiration
Important decisions often get stuck in subjective debates: one person prioritizes speed, another values quality, and neither knows what it would take to change the outcome. FlipPoint turns those invisible trade-offs into something people can explore, without a black-box recommendation engine.
What it does
FlipPoint is an interactive decision lab for comparing multiple options against weighted criteria. People can edit their decision, add or remove criteria and options, adjust importance sliders, and score each option from 1 to 5. Rankings and visual comparisons update immediately.
Its signature feature is the breakpoint. FlipPoint searches for the smallest change to any one criterion's weight (between 0 and 100) that would produce a different winner. Users can apply that change with one click and see the result reverse. The question becomes: Which assumption is actually decisive?
The prototype includes three sample decisions, local browser persistence, and an exportable Markdown decision brief. No login is needed, and no decision data is sent to a server.
Try the demo — 60-second walkthrough
- Open the live FlipPoint demo and start with Product launch.
- Lean beta initially leads with 4.11/5.
- Adjust a weight slider or 1–5 score button and watch the ranking change.
- Scroll to What would change your mind? In the original example, reducing Cost efficiency from 20 to 2 makes Research sprint the leader.
- Click Test this scenario, then Restore example to reset.
- Try Vendor selection or Career direction, or export a decision brief.
How it was built
The working prototype uses semantic HTML, custom responsive CSS, and JavaScript ES modules. A pure calculation module performs normalized weighted averages, deterministic ranking, and exhaustive single-variable sensitivity analysis. It checks every integer weight from 0 to 100, holding other inputs fixed. Ties do not count as a clear win.
There is no backend, account, AI inference API, or external data transfer. Browser storage retains the current decision locally; the Markdown export is generated client-side. Node.js built-in tests validate the calculation engine.
Challenges and iteration
The main challenge was ensuring that the breakpoint is both understandable and mathematically correct. A different option must strictly outperform the existing leader, and the recommended adjustment must be the smallest change to a single weight. Separating the engine from the interface made this testable.
What was actually tested: Seven automated checks pass, including an exhaustive minimal-breakpoint assertion, score calculation, proportional weights, input validation, zero-weight behavior, and exported content. On the deployed prototype, applying the breakpoint changed the winner; a new option survived a browser refresh; and Markdown export triggered a download. These were functional checks, not an external participant study.
Accomplishments
- A public, usable prototype with a distinctive core interaction.
- Seven passing automated tests.
- Three original synthetic demonstration scenarios.
- Responsive interface with semantic labels and reduced-motion support.
- Transparent scoring and a clear explanation of limitations.
What I learned
Decision support is more useful when it exposes the assumption that can reverse a recommendation. A transparent interactive model encourages a more focused discussion than a score in isolation.
What's next
Explore multi-variable sensitivity, evidence and confidence levels per rating, collaborative reviews, and formal user testing with product teams, including assistive-technology accessibility checks.
Tools, assets, and contribution disclosure
This original prototype was built during Silicon Valley Buildathon 2026 with ChatGPT AI-assisted implementation for UI code, logic, tests, editorial content, and example data. Samer Tallauze initiated the Buildathon entry and provided the project brief. Google Fonts supplies DM Sans and Manrope with system-font fallbacks. No third-party images, private datasets, or prebuilt project were reused. The sample scores are illustrative judgments, not objective predictions. No external users or human research participants have been claimed.
Method limitation: The breakpoint changes one integer weight at a time while holding other weights and ratings fixed. It does not test combinations of changes or establish statistical confidence.
