FastForward by Whack A Mole

Inspiration

AI can suggest faster code, but developers still need to check whether it is correct and whether different changes work together. We built FastForward to make that process measurable and transparent.

What it does

FastForward generates optimisations, tests correctness, benchmarks performance and highlights conflicting changes. Developers approve their preferred edits, then download a combined version that has been tested again.

How we built it

We combined a Python backend, a local AI model through LM Studio, and an HTML, CSS and JavaScript interface. Testing includes boundary cases, random inputs, exhaustive small cases and AI-generated tests targeted to specific changes.

Challenges we ran into

Handling unreliable AI responses, combining conflicting edits, measuring performance consistently and presenting statistics clearly.

Accomplishments that we're proud of

A complete optimisation-to-download workflow with targeted testing, explainable scores, conflict detection and human approval.

What we learned

Faster code is only useful if it preserves behaviour. Individual improvements may conflict, and passing tests provides evidence (not proof) of correctness.