# WebLens

- **Event:** [Built with Opus 4.7: a Claude Code hackathon](https://cerebralvalley.ai/e/built-with-4-7-hackathon)
- **When:** Apr 21 at 12:00 PM – Apr 27 at 2:00 AM (EDT)
- **Where:** Online
- **Team:** [hojjat rakhshani](https://cerebralvalley.ai/u/hrakhsha)
- **GitHub:** https://github.com/hrakhshani/generative-user-experience-hackathon-anthropic/tree/main
- **Demo video:** https://drive.google.com/file/d/1_h0x9lNJvaSwaibWgaq7QDuQxqj-GFgj/view?usp=sharing
- **Gallery:** https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery/212

User-Generative Experiences

What if you didn't have to wait for a web application to build the features you want?

Today, users are entirely dependent on digital product owners to shape their experience. Want to filter LinkedIn posts by topic? Compare hotels side-by-side on Booking.com? Merge identical announcements about a new Claude release on social media into one? You either wait or you go without.

I asked this question: what if users could build their own web experiences without writing a single line of code?

The proposed approach

I focused on web applications. This insight was simple but powerful: every web page can be decomposed into minimal, meaningful elements: posts, headers, search bars, cards. Once you can identify those elements you can assign behaviors to them.

I built a platform where users create an account, define reusable behaviors, and then at runtime drag and drop those behaviors onto any element of any web page. Every behavior is versioned and stored securely in a central database tied to the user's account. In this way, on any system when login into your account, all those behaviors are immediately accessible to you. For this hackathon, the latest Opus 4.7 model is used in conjunction with Claude Code and V0 to develop such a platform. During rune-time behaviors are created using Claude Code API. Given the limited time, the current implementation is tested against the following roles

Three Behaviors We Built

Filter: Assign this to a group of elements (like a LinkedIn feed) and it dynamically generates categories letting you then to filter and focus on only what's relevant to you. LinkedIn doesn't offer this natively.

Translate: Assign this to any element, select your target language and a translation button appears on every subsequent visit. The behavior persists no reconfiguring needed.

Summarize: Assign this to any content element and a summary button appears. One click, and you get the core of what's there.
Why Now

This is only possible today because of generative models like Claude. The intelligence needed to parse arbitrary web elements, generate dynamic UI, and produce meaningful functional outputs during run-time(e.g., translation, summarization, categorization)  is now accessible without a team of engineers behind it.

Browsers like Comet and platforms like Atlas are pointing in this direction. We think there's still a wide-open space at the user layer  and that's exactly where we're building.

---

Markdown version of https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery/212. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt