Kairos

The story of Kairos

You capture. Kairos decides.

It is 1 pm. You have a meeting until 2. Your list holds 47 tasks, and not one of them says “start with me”.

Ten of them have been sitting there for weeks. In every to-do tool I tried, what you do not do never rises to the top: new tasks pile on top, old ones sink. Every morning you have to re-sort the list in your head before you can work.

The pile is endless. The tool might as well sort it for you.

The observation

To-do tools manage a list, more or less finely. None of them sorts it. I had the idea of merging three worlds: agile development (a backlog you prioritise continuously), a to-do tool and the GTD method (capture everything, then process it).

The result: a list that evolves by itself over time. You take care of GTD, Kairos tells you what to start with. You can trust it enough to unload your mind, and put your attention on something other than managing to-dos.

Kairos Day view: the task to do now, then the tasks in order
The Day view, with example tasks: the task to do now, then the next ones in score order. No list to re-sort: the order is already done.

How it happened

  1. A script. The first version lived in a Linux terminal: a Python script that ranked my tasks with the WSJF score, and nothing else. It worked, and it changed my days.
  2. An interface. Then a real interface, served locally and started with systemd. The day I understood that the useful part was not the sorting but fitting tasks into the gaps of the calendar, the tool took its shape.
  3. Other people. Talking about it around me, I saw that the need was not only mine. A tool that only runs on my machine, with a service to install by hand, helped nobody else.
  4. Rebuilding everything. I rebuilt it all in Kotlin with Claude Code, to run everywhere: Android, Windows, Linux, macOS and the browser, with no account and no network.
Terminal: the kairos --capture command with a priority and points, then kairos --list, which shows the tasks sorted by descending score: fix the backup script (5.33), answer the supplier’s email (4.00), then the others
Reconstruction of the first version, before any git: two commands, tasks sorted by score. No duration, no deadline, no calendar. It is not a screenshot from back then: I recreated it from memory.

No AI: machine learning

The next ideas came with the wish to do machine learning and fine statistics on your own way of working. No generative AI, no chatbot, no cloud: pure algorithmics, on your device, that sharpens over time and helps you understand how you work.

The goal: the more you use Kairos, the better it estimates how long things will take you, and it always tells you which task to do now.

Why 14h55

14h55 (2:55 p.m.) is the project's code name.

According to a 2013 survey conducted by LondonOffices.com among more than 400 British employees, 2:55 p.m. is the time of day when workplace productivity is at its lowest. However, this finding is based on a self-reported survey, not on a peer-reviewed scientific study.

It is, however, part of a phenomenon better established by research: the post-lunch dip, a temporary decline in alertness and performance generally observed between 2:00 p.m. and 4:00 p.m. Research on circadian rhythms and cognitive fatigue shows that attention can decline after lunch, before picking up again in the late afternoon.

Sources : Daily Mail, « 14 h 55, l’heure où personne ne travaille ! » (2013)  ; Metro, « 14 h 55 : l’heure la moins productive de la journée au bureau » (2013)  ; Valdez et al., « Les rythmes circadiens et l’attention », Yale Journal of Biology and Medicine (2019)  ; Sievertsen, Gino & Piovesan, « La fatigue cognitive influence les performances des élèves aux tests standardisés », PNAS (2016) .

Who did what

Claude Code let me rebuild everything in Kotlin. The AI is my pen and a creation accelerator; the rest is mine: I set the course, planned, guided in detail, wrote the specifications, corrected, structured the code, set up the CI and the builds, drew the boundaries and brought the theory. Every change starts with a spec, before the code.

I answer every piece of feedback

Kairos is the first software I have published, and I answer every issue. A few examples:

What I do not hide

Kairos may not be for you if…

Mentioned on

You capture. Kairos decides.

Discover Kairos Open an issue