A prompt every fifteen minutes. Answering takes one tap, and it logs the time since it last asked, so the stretch you forgot to track is recovered instead of lost.
For Mac. Nothing leaves your machine, and it never sees your screen, your documents or the sites you visit.
Thirty-two frames, one per quarter hour, nine until five. Every frame you answer is exposed. Every frame you don't is time you cannot account for. A simulated minute passes every half second, so wait a few seconds and it will ask.
What have you been doing?
Whatever you pick is logged from 09:00 until now.
Three things happen, and only the first one asks anything of you.
It appears in the menu bar, or as a notification if you're in another app. One tap answers it. It never takes focus from what you're doing, one key dismisses it, and it stays silent outside the hours and days you choose. You can set the interval to 5, 10, 30 or 60 minutes instead.
This is the part that makes it work. Answering at 09:15 records 09:00 to 09:15. The quarter hour you spent head-down comes back to you, and you never have to have remembered anything. Miss a prompt and it asks again, still holding the whole interval.
Change a start time, add a session you forgot entirely, split one block into two, undo a delete. If you leave a timer running through lunch, it notices you were away and asks whether to keep the time, discard it, or log it as something else. An honest record is one you're allowed to correct.
The usual fix for forgetting is surveillance: the app reads your screen, your documents and the sites you visit, and infers the rest.
It works. It also means a piece of software holds a record of everything you looked at. That is not a trade everyone wants to make, and not something a coach can comfortably ask a client to install. This one has no idea what is on your screen. Everything it knows, you told it in one tap.
| Complete record | Reads your screen | Cost | |
|---|---|---|---|
| Manual timers | Only if you remember | No | £45–120 a year |
| Automatic trackers | Yes | Yes — apps, documents, URLs | £100+ a year |
| Quarter Past | Yes, by asking | No | £35, once |
Write down what you mean to do this week: a line each, with hours against them if you want hours. The app keeps the other half itself, and one button copies both to your clipboard for Claude or ChatGPT.
Everybody has the plan somewhere: a list, a calendar, three priorities agreed out loud with a coach. What nobody has is the second half, because working out where the hours really went is the tedious part, and doing it from memory quietly produces the week you would like to have had. That half is counted here rather than remembered.
Planning is optional and the record works without it. Nothing is scored, and there is no streak to keep: the app puts the two columns side by side and stops there. What to make of the difference is yours.
The brief counts before it says anything: how fragmented the days were, how much arrived in unbroken stretches, when in the day you actually worked. Then it tells the model plainly how complete the data is, so it hedges where it should. Hand a model five hundred raw rows instead and it spends its effort on arithmetic it is bad at, then reads meaning into noise.
The model is only what makes it quick. You could do the same comparison with these figures and a pen. You wouldn't, and neither did I.
The real shape of what gets copied, with illustrative figures.
## What I'd like from you
This is my own time-tracking data. Act as a thoughtful coach rather
than a summariser — I can read the totals myself.
1. Tell me what's actually happening, including the thing I
probably don't want to hear. Be direct.
2. Compare the plan against what actually happened. The next
section has both halves. Be straight with me about the gap,
including the parts of the plan nothing in the data can speak to.
3. Respect the data-quality section below. Where coverage is poor,
say so rather than reading meaning into noise.
Finish with at most three specific, testable changes I could make
next week.
## Plan against actual
This is what I said I'd do in the week beginning 7 September, set
against what I actually did.
| What I meant to do | Planned | Actual | Difference |
| Riverside | 12h | 6h 30m | −5h 30m |
| Keep admin down | 3h | 6h 30m | +3h 30m |
| Finish the Q3 tender | no target | not tracked separately | |
| Planned in total | 15h |
| Spent on work I had planned | 13h, 91% of tracked time |
| Spent on everything else | 1h 20m, 9% of tracked time |
Before you read anything into the difference:
· One of the 3 lines names no project, so nothing in this data
says whether it happened. Ask me rather than assuming.
## At a glance
| Total tracked | 83h 4m (83.06 h) |
| Sessions per day | 4.0 |
| Project switches per day | 2.3 (0.6 per tracked hour) |
| Unbroken focus blocks ≥1h| 42 — 58h 52m, 71% of tracked time |
### When the work happened
09:00 ██████████████████████·· 13h 18m
11:00 ████████████████████████ 14h 15m
14:00 ██████████████·········· 8h 21m
16:00 ██████·················· 3h 24m
## How much to trust this
| Measured live | 34% — a timer was running as it happened |
| Reconstructed | 66% — by hand, or filled in when prompted |
| Days with any data | 21 of 30 |
About 1,700 words for a month of tracking. Small enough to paste anywhere.
I didn't, properly. I'd remember on the Tuesday, forget by the Thursday, and turn up having reconstructed the fortnight in the car.
Which wasted half the session, mine and hers. We'd spend the first twenty minutes assembling a picture of my week instead of doing anything useful with it. Every tracker I tried either needed me to remember, or wanted to read my screen to find out.
So this asks instead. It's the tool I wanted for that conversation, and I use it every working day for exactly that.
If you're a coach: ask your clients to bring their analysis to the next session and see what the conversation turns into. Get in touch and tell me how it goes.
It isn't on sale yet: the app is in final testing while Apple verifies the company. This is what it will cost.
For comparison: Tyme is about £45 a year, Timing about £100. Those are good tools. They're just rented.
One message when it's ready. Nothing before that, and nothing else afterwards.
One email when it ships. Nothing else, and one word back takes you off again.
It would be if it interrupted. The prompt is a card you can ignore, it never takes focus, one key dismisses it, and it's silent outside the hours and days you set. You can also choose 30 or 60 minutes. The honest answer is that a prompt you answer in one tap is less annoying than reconstructing a fortnight from memory.
No. The record works on its own, and most of what the app does never touches the plan. If you do write one, it is a line per intention, and hours are optional on each. Nothing is scored and there is no streak to break. The app puts what you meant next to what happened and leaves the conclusion to you.
No. It has no idea what's on your screen, what you're typing or which sites you visit. The only thing it reads is whether the machine has been touched recently, which is how it notices you left a timer running through lunch. Everything else, you told it.
Nothing — it's one payment, so there's nothing to lapse. And even if a licence can't be checked, the app still opens, still shows your whole history, still runs the analysis and still exports everything. Only recording new time stops. Your time is yours.
On your Mac, in a file you can back up, export and read. There's no account and no server. That also means keeping a backup is your job, and the app makes it one click.
Any Mac from macOS 10.15 onwards, Intel or Apple Silicon. The whole app is about 8MB.