Should I vibe code
Run visual project boards with recurring tasks, timelines, and lightweight reporting
The board is an afternoon. Two people dragging the same card at the same moment is the product.
?
Their verdict, the Pro price and the build-time estimate come from their entry, MIT-licensed. Checked 2026-08-04.
?
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
Boards, columns, cards, drag and drop: the demo is an afternoon and it will look right. What separates it from MeisterTask is not features, it is two people touching the same board at the same second. Once there are two you need presence, conflict resolution, an ordering scheme that survives simultaneous drags, and notifications that are neither silent nor a firehose — and none of those announce themselves as missing. They show up as a colleague who has quietly stopped trusting the board. MeisterTask's own free tier already covers three projects with unlimited members, which is a hard thing to beat by writing code, and the paid features are precisely the ones a homemade board fakes worst: recurring tasks, automations that fire on section changes, and permissions. Nothing catastrophic lives here. What lives here is a team drifting back to Slack messages and a spreadsheet, which is worse than a crash because nobody files it as a bug.
What actually breaks
not "if". the specific failures.
- Card ordering, the first time two people reorder the same column in the same second — integer positions collide and somebody's drag silently loses
- Recurring tasks, which are a recurrence rule and a timezone, and which will either spawn duplicates across a daylight-saving boundary or skip a week without telling anybody
- Notifications, which land either too quiet to trust or loud enough that everyone mutes them, and there is no third setting you will stumble into
- Permissions, retrofitted onto a data model that assumed everyone could see everything, usually the week an external collaborator joins
- Comments and attachments, which is where the actual decisions end up living and which nobody remembers to include in the export
- Flaky mobile connections, where an optimistic update that was never confirmed leaves one person looking at a board nobody else has
- Archive and delete, and the afternoon somebody needs a task from six months ago that a cleanup job tidied away
- Search, once there are ten thousand cards and a LIKE query stops being a search feature
Is that you?
the verdict is a default, not a law
- It is one person's board, or a couple of people who are never in it at the same time
- The workflow is genuinely unusual — a pipeline with stages no commercial tool models — and that shape is the entire reason to build
- Losing the board would cost you an afternoon of retyping rather than a project's history
- You are treating it as a front end over data you already own somewhere else
- More than about three people rely on it to know what they are supposed to be doing today
- Anybody outside the team needs access to part of it and not the rest
- Recurring work is meant to appear without somebody remembering to create it
- The comment threads are where decisions are recorded
If you build it anyway
the checklist, then the prompt that enforces it
- Use fractional or lexicographic ordering keys for cards from the first commit. Integer positions with a reindex step are the bug that makes a shared board feel haunted, and retrofitting the fix means rewriting every list.
- Decide the concurrency story before the UI: last-write-wins on whole cards is fine for a small team if you say so out loud, and catastrophic if you assumed field-level merging you never built.
- Compute recurrences in UTC with an explicit timezone per project, and materialise the next occurrence when the previous one completes rather than on a schedule that can be missed.
- Build export before import. CSV or JSON of cards, columns, comments, attachments and completion history, runnable by anyone, from week one.
- Model permissions as a real thing early even if everyone is an admin today, because adding a visibility boundary later means auditing every query you have written.
- Keep deletions soft and recoverable for a fixed window, and make archive genuinely different from delete.
- Put attachments in object storage rather than the database, and decide the retention rule before somebody uploads a two-gigabyte video.
I am building a shared project board — projects, sections, cards, assignees, recurring tasks and notifications — for a small team. The demo is easy; the things that make it usable by a second person are not. Order the work so those come first, and push back when I ask for views instead. 1. Start with card ordering. Use fractional or lexicographic keys, never integer positions with a reindex, and write the test for two clients reordering the same column concurrently. This is the decision that is most painful to change later. 2. Then state the concurrency model in the README before writing any UI: what happens when two people edit one card at once, and whether resolution is per card or per field. If the honest answer is last write wins, say so rather than implying a merge that does not exist. 3. Then export. CSV and JSON covering cards, sections, comments, assignees, attachments and completion history, runnable from one command with no admin account. Before any charts, before any timeline view. 4. Recurrences in UTC with a per-project timezone, materialised when the previous occurrence completes. Write the daylight-saving test first, and make a skipped occurrence visible rather than silent. 5. Soft-delete everything with a recovery window, and keep archive distinct from delete. Somebody will empty a column by accident in week two. 6. Model permissions now, even though everyone is an admin today: every query filters by an explicit access rule rather than assuming the whole team can see everything. 7. Attachments go to object storage with a size cap and a retention rule, never into the database. 8. Notifications are batched and per-project by default, with a digest option. An unmuteable firehose gets the whole tool muted. 9. Out of scope, and say so rather than stubbing it: real-time presence, Gantt and dependency scheduling, guest access, and automations. MeisterTask is 9 USD per user per month and its free tier already covers three projects.
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
Basically at the second user. Nine dollars per person per month is the wrong comparison anyway, because the free tier gives you three projects and unlimited members — the honest question is whether your workflow genuinely does not fit a board, and it usually does. Pay when you need recurring tasks, automations or permissions, which are exactly the three features a homemade version imitates badly. Build it yourself only when the shape of the work is genuinely unusual, and even then build the view and let something else own the data.
$9/mo is cheaper than your weekend.
Export is the whole exit and it has to include the parts people forget: comments, attachments and the completion history, not just open cards. Every commercial tool imports a flat CSV of tasks, so getting the cards out is easy and getting the conversation out is not — and the conversation is where the reasoning lives. If the board ever becomes the team's memory rather than its to-do list, run the export on a schedule and check that it restores, because the failure you are protecting against is not a crash, it is quietly outgrowing the thing.
Actively developed open-source project and issue tracking with boards, cycles and modules.
Self-hosted Trello-style board tool, now explicitly unmaintained and looking for a maintainer — a fair preview of what adopting one of these costs.
Questions
Why does card ordering keep coming up as the hard part?
Because it is the one piece of state that every user changes constantly and that has no natural conflict resolution. Integer positions require renumbering the column on every move, which means two simultaneous drags produce a board that is subtly different for each person, and that is exactly the bug that makes a team stop trusting a tool. Fractional or lexicographic keys make it a non-issue and cost nothing on day one — and rewriting every list to add them later is genuinely miserable.
MeisterTask has a free tier. Is building it ever rational?
Only when the shape of the work does not fit a board — a pipeline with stages, gates and rules that no commercial tool models, where the tool is really a domain application that happens to look like cards. If you can describe what you need as "Trello but", you are describing a product that exists. The free tier covering three projects with unlimited members is a high bar to clear with a weekend.
What breaks first in practice?
Notifications, and not in a dramatic way. You will either send too few, so people stop looking at the board because it never tells them anything, or too many, so they mute it and stop looking at the board. Both end in the same place, which is the team going back to Slack and the board becoming a museum. Getting that dial right is unglamorous product work, and it is a large part of what a subscription is actually buying.
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.
Project management tools are easy to build and impossible to get a team to adopt.
Configurable boards for one team is achievable. Configurable boards for any team is a platform.
Tasks, docs, goals, dashboards and automations. That is five products and you want all five.
last reviewed 2026-08-05 · verdict is editorial and unsponsored · shared entry data from canivibecodeit under MIT · not legal advice