Inspiration

Project planning rarely has one obvious answer. Most tools push users toward a single plan before they have explored the tradeoffs. We built Parallel to help people consider multiple strategies before committing.

What it does

Parallel turns a plain-language project idea into three distinct routes:

  • Cautious — prioritizes lower cost and risk
  • Baseline — offers a practical, balanced approach
  • Ambitious — aims for greater reach and impact

Users can compare costs, capacity, and risks; edit or fork any route; and merge the strongest decisions into one final plan without changing the original.

How we built it

We built Parallel with Next.js and React, using Google Gemini through the Vercel AI SDK. Gemini first extracts a structured brief from the user’s description, then creates three comparable plans. Zod validates the generated data, while a deterministic scoring system checks that the routes are meaningfully different and respect the project’s constraints.

Projects are saved locally in the browser. The server also applies request validation and rate limiting, with optional Upstash Redis support for distributed deployments.

Challenges we faced

The hardest part was making AI-generated plans different while keeping them realistic and directly comparable. We addressed this with structured generation, strict schemas, quality scoring, and a correction pass for weak results.

Designing a three-way merge system was another challenge because it needed to preserve plan ancestry, identify real conflicts, and let users resolve them clearly.

What we learned

We learned that reliable AI products need more than a strong prompt. Schema validation, deterministic checks, clear constraints, and thoughtful error handling are essential.

We also learned how concepts from version control—branching, ancestry, comparison, and merging—can make decision-making easier to understand.

What's next

Next, we want to add collaborative planning, shareable projects, richer scenario analysis, historical comparisons, and smarter recommendations based on each user’s priorities.

Built With

Share this project:

Updates

Submission history