Inspiration

We noticed that traditional writing tools just… don’t get how some of us think. ADHD and neurodivergent minds often freeze mid-sentence, jump ahead, or make the same typo over and over. That’s not a bug — it’s just how our brains work. So why should current editing tools interrupt that rhythm? How often have you misspelled something and
1.) autocorrect didn’t even get it right, and
2.) you completely lost your train of thought? </3

We built ThinkFast to help people express their ideas freely, without being held back by rigid, overstimulating editing systems.

What It Does

ThinkFast is a personalized autocorrect web application that lets you type freely — without getting overwhelmed by glaring error signals.
You can choose whether you want automatic corrections turned on or off, then write directly into the text box. When you're ready, view your corrected writing without having your thought process interrupted.

Want to see what you commonly mess up? Just click "View Insights" to get a pop-up with your most frequent spelling mistakes — so you can learn and improve over time.

How We Built It

We used Gramformer, a language correction model by Prithiviraj Damodaran, as the foundation of our Python backend. Gramformer enables smarter, context-aware corrections and allows for deeper personalization as users interact with the tool.

We connected our backend to a frontend built in HTML using FastAPI, enabling real-time, low-latency communication between the interface and the correction engine.

To keep the user experience ADHD-friendly, we intentionally:

Removed distracting red underlines
Added a toggle for live autocorrection
Built an "Insights" pop-up that tracks and visualizes your most common writing habits and misspellings

Challenges We Ran Into

We ran into many technical complications — especially when connecting our model to the user interface using FastAPI and HTML.
Some key challenges included:

Implementing per-word live autocorrection
Handling visual complications with the Insights dashboard
Balancing model training between our goals and Gramformer's defaults

In the end, we were able to bring our model and interface to a solid, user-friendly state.

Accomplishments We're Proud Of

We're especially proud of:

Developing a working web application that connects multiple components into a seamless user experience
Creating something that actually does what we intended — and is meaningful
Identifying a real-world problem and designing a solution that’s inclusive, accessible, and useful for many

We’re excited by the potential for this to help people express themselves more freely in everyday life.

What We Learned

From a conceptual standpoint, we learned:

How to identify nuanced needs within our community
How to build with accessibility and neurodivergence in mind
How to prioritize inclusivity and user experience over perfection

Technically, we learned:

A ton about APIs and full-stack development
How to integrate ML models with web frontends
The complexities of timing, UI state, and error handling in live correction workflows

What's Next for ThinkFast

Next steps we’re excited about:

Improving the model’s accuracy and speed
Making the user interface more interactive and customizable
Building faster, more seamless corrections
Expanding support for students and anyone who wants to express themselves more freely, without friction

We’d love to grow ThinkFast into a tool people actually rely on — not just to fix mistakes, but to think more confidently.