Curio is an internal research library I designed and developed at Framestore. New AI tools were being shared across messages and group chats, but there was no common way to organise them or find a useful link again a few weeks later. The project grew outside a client production, with the aim of making that research accessible to the teams.

Contributing needed to be as quick as posting a link in a chat. Users submit a URL, and Curio extracts the content to generate a title, summary, tags and thumbnail. I also built a Chrome extension so people could save a reference directly from the page they were reading, without opening the platform or filling in a form.

Each card shows its tags, whether the tool is open source or proprietary, and how old the information is. I added the age indicator after feedback from my supervisor, so producers could quickly judge whether a reference was still relevant. Contributors can correct summaries and categories, while other users have read-only access.

The first version ran locally before becoming a web service using company Google accounts. I handled the architecture, interface, extension, deployment and documentation, then worked through login, display, stability and performance issues reported after launch.

Four months after launch, Curio had 47 accounts and 233 indexed links, with four to eight active contributors per week. One producer used it to identify local character motion-transfer options and prepare a discussion with the department before involving the team.