"What did I learn about machine learning this week?"

Daylens is a local-first memory system for your laptop. It watches what you do, enriches it with Claude, and lets you ask anything in plain language.

Daylens watches without integrations, plug-ins, or permissions you'll regret.
The data you generate every day should belong to you. So it does.
Sessions watched this week
Apps in active orbit
Data shared with anyone else
Four queries, four real answers from real data on a real laptop. No mocks. No edits.
The privacy model is not a feature. It is the constraint that makes the rest of the product trustworthy.
A single SQLite file on your laptop. No cloud. No account.
Free, forever, on every Mac.
Open source. CBC Spring 2026 Hackathon submission.
Daylens is designed around the inverse of surveillance. You see what is captured. You decide what your AI can read. The product is self-knowledge under your control.
Substantive use of the API. Not a chatbot wrapped around your schedule.
A background job sends every new browser session to Claude with a two-sentence, topic-tagged prompt. Results land in a local SQLite table the chat can search. Built this week.
Claude with SQLite tables and the content index as tools. Deterministic router handles common shapes first. searchContentByTopic returns what you actually consumed.
Opt-in MCP exposes Daylens to Claude Desktop, Cursor, and Claude Code. Your work history becomes context for any agent. Off by default, revocable from Settings.
Every byte lives in a single SQLite file on your laptop. The only thing that leaves the device is the question you choose to ask.
A three-minute walkthrough against live data. Same four queries. Real answers.
demo · uploading
Open source. Local-first. Available now for macOS. Windows and Linux are next.
Source on GitHub.