Most bookmark products fail at the exact moment people need them most: after a few hundred saved links, everything turns into a private landfill of vague titles, nested folders, and dead URLs. The link survives, but the reason for saving it disappears. Comparing sources means reopening every page and reading from scratch. I built Logged to change that outcome: a personal archive that stores the actual content and context behind each bookmark, then lets users query the whole thing through AI.

Logged is deliberately a single-user product, not a team workspace. That constraint mattered. The job was not collaboration, permissions, or shared knowledge bases. The job was making one person’s messy, long-running archive useful again without asking them to rebuild it from zero. So I kept the user’s existing folder structure exactly as it was and layered enrichment on top. That removed the biggest adoption barrier: nobody reorganizes 800 bookmarks just to test a new tool.

From there, every feature fixed a specific failure in traditional bookmarking. Each saved link can carry a memory note, so users stop asking “why did I save this?” AI-generated tags and user tags sit on top of folders, which means retrieval no longer depends on remembering a page title that never described the value of the page. Dead link detection automatically moves broken URLs into a separate list, which keeps the archive current instead of quietly decaying.

The most important product decision was also the most expensive one: Logged does not just save a pointer. It stores full metadata and up to 100,000 words of scraped content from each link. That changed the archive from a list of references into a searchable corpus. It costs more in storage and processing, and it is the reason the product works. Without that content layer, chat is decoration. With it, the archive becomes answerable.

That is where Atlas comes in. I designed Atlas as the interface to the archive, not a chatbot pasted onto a bookmark list. Users can ask questions across saved sources, compare multiple sites, or retrieve something they only half remember saving, all without opening the original links. That turns research from a repetitive tab-opening exercise into a single query. For researchers compiling sources on one topic, that means instant comparison across saved material. For self-directed learners with courses, docs, and tutorials spread across dozens of tabs, it means the material stays usable after the tab is closed.

We built Logged in 4 weeks with custom product design by a unicorn designer, development in Claude Code, OpenAI agents for the internal research agent, and Supabase for database and storage. The short timeline forced sharp product decisions early, and the product shipped live with a Chrome extension, full brand, and complete product design.

The outcome is simple and concrete: archives hold thousands of bookmarks and stay navigable, where traditional folder systems break down after a few dozen. Users do not browse a link cemetery anymore. They talk to a memory tool.