Where did today actually go?
You worked all day. You shipped something. But if a client asks for hours, or you ask yourself what took the afternoon, you're guessing.
Manual timers only know what you remembered to start. App trackers know you were "in Chrome for two hours", not which client or which task. Both leave you rebuilding the day from memory.
A timer measures your intentions. Your day is what actually happened. ( The work you never see on a to-do list is usually where the time went.)
How does AI time tracking work in Didon?
Add your context.
List your projects and the categories you care about, such as Coding, Marketing or Client calls. This takes a minute.
Work as usual.
Didon starts with your Mac and captures a screenshot every 30 seconds in the background. It pauses when your Mac sleeps.
Local AI reads the screen.
An on-device model looks at window titles, file names and visible text, and matches each moment to one of your projects. Anything that doesn't match is labeled "Other".
Get your journal.
At the end of the day you have time per project, a category breakdown and a written summary of what you did.
What you get
Context, not app names
Didon sees the ticket, the file and the client name on screen. Your log says "Refining login form validation for Acme", not "Cursor, 47 min".
A daily work journal
Every day becomes a structured entry: time on each project, a category split and a short written summary. It's useful for standups, client updates or your own review.
Ask your logs anything
Type a question in plain language, such as "What did I work on for Acme last week?", and get an answer from your own history.
CSV export for billing
Daily logs save automatically and export to CSV. Send them to a client, drop them into an invoice or keep them for your records.
Zero start, zero stop
Auto-launch tracks the first minute of your day. Auto-pause on sleep keeps idle time out of your hours. There's nothing to remember.
Reports that show patterns
Weekly and monthly breakdowns by project and activity type show where your hours go across a week, not just today.
Can an AI time tracker be private?
Yes, if the AI runs on your machine. Didon's screen analysis uses a local vision model (Qwen-3-VL:2b) on your Mac.
- Screenshots are analysed on-device and never uploaded to our servers.
- Work logs and reports are saved locally in
~/Documents/Didon/. - No one sees your activity unless you export it and send it yourself.
This is tracking for your own awareness, not monitoring. You own the data, and you decide what leaves the machine.
Who is Didon for?
Developers.
See how much of the week was real build time and how much was reviews, meetings and context switches.
Founders and indie hackers.
When you run three projects at once, see which one actually got your hours.
Freelancers and agencies.
Bill from a log written as you worked, not from memory on Friday. ( How work journals keep freelancers organized)
Remote workers.
Get a clear record of your day without a manager watching over your shoulder.
How is this different from a manual timer?
| Manual timer | App-usage tracker | Didon | |
|---|---|---|---|
| Starting and stopping | You do it, every task | Automatic | Automatic |
| What it records | What you remembered to log | App and website names | The task and project on screen |
| Matches work to your projects | By hand | Rules or manual sorting | AI, from your project list |
| Written daily summary | No | Rarely | Yes |
| Where the analysis runs | n/a | Usually in the cloud | On your Mac |
| Idle time | Counted if you forget to stop | Varies | Paused on sleep |
Read the longer breakdown: AI productivity tracker vs manual work tracking .
Simple pricing
- Free: automatic AI tracking, daily logs and project categories, with 1 week of history.
- AI Work Tracker, €9/month: unlimited CSV exports, productivity metrics, and weekly and monthly breakdowns.
- AI Work Manager, €20/month: everything in Work Tracker, plus an AI manager on Telegram for feedback and accountability.