Lock-In

Inspiration

There are a lot of mobile phone productivity apps to help track screen time, create overlays ect… However, the majority of work is done on laptops and browsers such as chrome. We wanted to create a productivity AI assistant that created a buddy like experience to help keep us on track. Other existing tools such as website blockers are too rigid and block more than what’s necessary instead of adapting to real-time needs. We wanted to build a smarter, AI-powered tool that helps users stay focused without completely cutting them off from resources like youtube. This led us to Lock-In, a dynamic Chrome extension that keeps users on track while allowing flexibility.

What It Does

Lock-In is an AI-powered Chrome extension designed to enhance productivity by detecting distractions in real-time. Users start a session by inputting their current task, and Lock-In actively monitors the sites they visit. If a site is unrelated to their task, the AI sends a real-time notification to refocus. Once the session ends, Lock-In generates an analytical report summarizing productivity levels and categorizing website activity. This way, users gain insights into their habits and can make data-driven improvements to their workflow.

How We Built It

The frontend is built using React as a Chrome extension. The backend relies on AWS Lambda for serverless processing, API endpoints to handle user requests, and DynamoDB to store session data and generate reports. To determine the nature of visited websites, we tested multiple LLMs (large language models) and ultimately chose ChatGPT for its superior accuracy and speed in classifying web content. The AI integration allows Lock-In to distinguish between productive and distracting websites in real time. Challenges We Ran Into

Integrating an LLM API in a Chrome extension presented several technical hurdles, from handling API rate limits to optimizing response times for real-time feedback. Additionally, fine-tuning the AI to classify websites into productive and non-productive categories while avoiding false positives took extensive testing.

Developing the frontend in chrome extension was a journey into the unknown for the team. We faced many obstacles in the differences in web vs chrome extension development. Specifically, the limitations of chrome extensions and how it works primarily in vanilla html, css and javascript pages. Creating a user-friendly analytical dashboard within the extension was another challenge, requiring us to balance performance with detailed reporting.

Accomplishments That We're Proud Of

We successfully built an AI-powered productivity tool that adapts to users' needs instead of enforcing a one-size-fits-all approach. Our real-time distraction detection system and post-session analytics provide meaningful insights, setting Lock-In apart from traditional website blockers. We’re especially proud of how we optimized LLM queries to ensure fast and reliable site classification without slowing down the user experience.

What We Learned

Building Lock-In reinforced our knowledge of Chrome extension development, AWS serverless architecture, and AI model integration. We gained hands-on experience with optimizing API calls, processing real-time data, and designing intuitive user interfaces. Additionally, testing different LLMs gave us valuable insights into how AI models interpret web content and how to fine-tune prompts for better accuracy. We’ve gained many insights into the difficulties of chrome extension development and how that differs from other platforms we’re more familiar with. This includes both the limitations and native benefits of chrome extensions.

What's Next for Lock-In

We plan to expand Lock-In by adding customizable distraction thresholds, allowing users to set different levels of strictness based on their focus needs. We’re also exploring more advanced AI models for even better website classification and gamification features like productivity streaks and rewards. Additional productivity tools to help users stay focussed would include blurring out distracting content on websites real-time and potentially incorporating

Built With

• amazon-web-services
• chrome-extension-api
• dynamodb
• javascript
• openai
• react