Should I vibe code
Track time with projects and reminders using software-only controls
The timesheet is a Saturday. The eight-sided cube on your desk is the actual product, and it is not made of software.
?
Their verdict, the Personal price and the build-time estimate come from their entry, MIT-licensed. Checked 2026-08-05.
?
Our verdict, the regret score and everything below it. Editorial and unsponsored — nobody can pay to be moved.
The honest answer
why the verdict is what it is
Strip the cube away and what is left is one of the calmest builds on this site: a table of time entries, a list of projects, a weekly total, and a CSV button. That is a Saturday, it is yours, it runs on one machine, and the worst outcome is losing a history nobody else wanted. The reason Timeular costs money is the thing you cannot generate — an eight-sided Bluetooth cube that sits next to your keyboard, and the small physical ritual of flipping it, which is what actually gets time tracked. Two facts worth knowing before you decide. The company rebranded: timeular.com now redirects to early.app, the product is EARLY, and the Tracker will not work without an active EARLY subscription, so buying the hardware and pointing it at your own software is not an option they offer. And if you build your own puck anyway — a BLE orientation sensor is a genuinely fun project — understand that the failure mode of Bluetooth is silence. The cube keeps flipping, the app keeps looking fine, and on Friday you have four hours logged for a week you worked forty.
What actually breaks
not "if". the specific failures.
- The Bluetooth link, which drops without telling you — the cube still flips, the app still draws a running timer, and the pairing quietly ended forty minutes ago
- Orientation debounce, because a cube nudged while you reach for a mug generates three side changes in two seconds and three entries you have to delete by hand
- The empty side, which is the whole idle story: forget to flip back and Tuesday afternoon is nine hours of one client
- OS Bluetooth permission prompts, reset by macOS and Windows updates, after which the tracker records nothing until you notice
- Battery, six months of it, and the day it runs out is a day with no timesheet and no warning
- Rounding, which sounds trivial until the six-minute increments you invoice from disagree with the raw seconds you stored
- The single SQLite file holding a year of billable hours, on one laptop, backed up by nothing
- Retroactive edits, if corrections overwrite the original samples rather than sitting beside them — then your memory of Thursday becomes the only record of Thursday
Is that you?
the verdict is a default, not a law
- It runs on your own machine and tracks nobody but you
- The entries are manual or come from a device you own, and nothing syncs to a server you also have to keep alive
- The output is a weekly total you read, not an invoice that goes out unreviewed
- You are relaxed about the whole database being lost, or you export it somewhere on a schedule
- Anyone else's hours land in it — a team timesheet is a payroll input and a completely different entry
- You bill clients straight from the totals with no review step between the raw samples and the invoice
- You were counting on the official Tracker working with your software, because it will not without an EARLY subscription
- The project names in it are client names you would not want in a synced folder
If you build it anyway
the checklist, then the prompt that enforces it
- Build the manual timer first and use it for two weeks. Most of the value is in the report, not the input device, and you will find out whether you actually track time before you order a dev board.
- Treat a Bluetooth disconnect as an error, not a pause. Show the connection state permanently, and write a gap marker into the data rather than letting a live timer keep running against a device that left.
- Debounce side changes by a couple of seconds and ignore transitions shorter than that. Physical objects get knocked.
- Store raw start and stop events immutably, and hold every manual correction as a separate layer that references them. Never overwrite an observation with a memory.
- Put an idle detector on top regardless of hardware: prompt when a single entry passes a threshold you set, because the forgotten flip is the dominant error in this design.
- Export to CSV on a schedule from day one, to a directory you back up. A time log is only worth anything cumulatively.
- Keep the store local. If you add sync, remember the file is a list of who you worked for and when, which is more revealing than it looks.
I am building a personal time tracker, optionally driven by a Bluetooth
orientation sensor I flip on my desk. The dataset matters more than the
hardware, and the hardware fails silently. Build in this order.
1. Start with the data model and the manual timer. No device code at all in
the first version — I want to know whether I use this before we talk to a
radio.
2. Raw events (start, stop, project, source) are append-only and never
edited. My corrections go in a separate table that references them, so the
original observation always survives.
3. Local SQLite, no account, no telemetry, no server. If I later ask for sync,
stop and tell me the file is a record of which clients I worked for and
when.
4. Write CSV and JSON export, plus a scheduled dump to a folder I choose,
before you write a single chart.
5. Only now the device. Treat disconnection as a first-class error state:
surface the connection status permanently, and on disconnect close the open
entry and write an explicit gap record. Never let a timer keep counting
against a device that is gone.
6. Debounce orientation changes — ignore any side that is held for less than
two seconds, and never create an entry from a transition burst.
7. Reserve one side as "not tracking" and add an idle rule independent of the
hardware: if a single entry exceeds a duration I configure, prompt me
rather than silently logging a nine-hour block.
8. Round only at report time, from the stored seconds. Do not round on write.
9. Fail loudly on permissions. Check Bluetooth authorisation on every launch
and show a persistent warning, because a week of quiet zeros is the worst
outcome here.
10. Out of scope, and say so rather than stubbing it: tracking anyone else,
team dashboards, invoicing, and anything that talks to the official
Timeular/EARLY Tracker, which needs their subscription to function.That one keeps you out of trouble. For the prompt that actually builds it, canivibecodeit.com has one.
their build prompt ↗Or don’t build it
the boring option, and the way back out
Buy it for the cube. Nine dollars a month for Personal, or fourteen for Personal Pro if you bill hours, is not buying you a database — it is buying an object that makes you actually record time, plus the app that already knows what to do when its radio drops. If the flipping ritual is what appeals and you have no interest in debugging a BLE reconnection loop, this is a clean purchase. If you only wanted the report, build the report.
$9/mo is cheaper than your weekend.
Everything of value is start, end, project and note, which is exactly the shape Toggl, Harvest, Clockify and Kimai import. Keep those five columns clean and leaving is a CSV away. The part that does not travel is the hardware: if you built your own puck, its firmware and pairing are yours alone, and the official Tracker will not follow you either — it requires an active EARLY subscription. So plan the exit as software-only, and keep the cube in the drawer with the other dev boards.
Mature open-source time tracker with projects, rates, reports and exports, and a sensible schema to copy even if you never run it.
Open-source desktop task and time tracker with local-first storage and issue-tracker integrations.
Open-source automatic tracker if you decide the cube was really about not having to remember at all.
Questions
Wait, is this still called Timeular?
Not really. timeular.com issues a permanent redirect to early.app, the product is branded EARLY, and the site footer credits TimeTac. The physical device is the same object — their own tracker page says so explicitly, noting that "in the past, the Tracker was called the Timeular Tracker, so you may still come across this name online." Nothing about the build advice changes, but if you go looking for the pricing page you were comparing against, it moved.
Can I buy the cube and use it with my own app?
Not as a supported path. EARLY sells the Tracker separately from the subscription but states that the Tracker requires an active subscription to use, so the hardware is not a standalone peripheral you can adopt. If you want a physical trigger under your own software, the honest route is building one — a BLE dev board with an accelerometer is a weekend and a soldering iron — and accepting that you now own the firmware too.
Is the software half really as easy as the other time-tracking entries here?
Easier, if anything. Timing and Rize have to observe your machine, which means accessibility permissions, browser URL extraction and an exclusion list before the first row is written. A flip-a-cube tracker records only what you deliberately tell it, so the sensitive-capture problem largely disappears and you are left with a small event log and a report. The interesting engineering has moved into the radio.
What actually goes wrong first?
The forgotten flip, every time. Bluetooth dropping at least leaves an obvious hole once you build the gap marker. A side left on "Client A" over lunch and a meeting produces a plausible, wrong nine-hour entry that you will not question three weeks later when you invoice from it. Build the duration threshold prompt early; it is worth more than the hardware.
Every week, someone ships something they shouldn’t have.
New verdicts, the worst thing that landed in the trap, and the occasional incident report. No other email, ever.
An automatic tracker is a minute-by-minute diary of your life in a SQLite file. Don't put it in Dropbox.
Tracking yourself is a Saturday project. Tracking your colleagues is a surveillance system with its own case law.
Wayland refuses to tell you which window is focused. That refusal is most of what you're paying for.
last reviewed 2026-08-05 · verdict is editorial and unsponsored · shared entry data from canivibecodeit under MIT · not legal advice