Lyubi Daily

The problem

Alexander Lyubishchev, a Soviet entomologist, recorded how he spent his time in five-minute resolution for 56 years, then used the record to plan the next year against what the last one actually cost. Not a productivity system — an accounting system, applied to the one account nobody reconciles.

The idea stuck with me for an obvious reason. I spent five years making sure every ledger balanced, and had no idea where my own hours went.

Existing time trackers are built for billing clients. They ask what project this is for. I wanted the opposite: start a timer, name the activity, and let the summary tell me the truth at the end of the day.

How it works

Deliberately small. Pick an activity, start, stop, optionally add a note. Two views on the same data:

• Summary groups by activity and shows total time per category, each card tinted with that activity’s own colour.
• Timeline lists the day in order, which is the view that actually stings — the gaps are as informative as the entries.

Activities are user-defined: name, emoji, and a colour from ten presets or a custom value that gets normalised on input. An “Edit Activities” panel edits all of them at once with live colour preview, because editing them one at a time was the first thing that annoyed me about my own app.

Working with AI on this

This was the project where I learned to work in versioned stages rather than one long conversation.

The rule I set: when a stage is finished and I have confirmed it works, the model gives me a commit message and appends a version note to the README saying what changed. That is why the README carries a “v1” section listing exactly what shipped. It sounds bureaucratic for a personal project. It is the difference between a codebase I can return to after two months and one I have to re-understand from scratch.

The second lesson was about scope. My first instinct was to ask for tags, filters, weekly rollups, and export — all at once. What came back worked and I did not understand any of it. Rebuilding it one stage at a time took longer and left me able to change it, which is the entire point of building your own tools.

What I would change

The data lives in the browser, which means it does not survive clearing site data. For a tool whose value compounds over years, that is the wrong trade. If I return to it, local-first storage with a real export path comes before any new feature — the same conclusion I reached independently while building Chalkbook.