Why Building WordFloor, a Private Writing App, Takes So Long
[]We’ve been building WordFloor for months. It is not ready yet.
Not in the usual way. There has been no sprint planning, no whiteboard covered in sticky notes, and no product manager asking whether we’ve considered adding AI.
We’ve been building it slowly, with a lot of arguments about the right way to do things. Then we’ve argued about whether the decision we’d made was actually right.
The first version was terrible. The second was slightly less terrible. The third almost worked. The fourth is still in progress, but it is getting close.
The hard part is not the technology. It is the restraint.
What WordFloor is
WordFloor is a private daily writing space built around one idea: effort can outlive content.
I explained that distinction in the first post about WordFloor.
You set a daily word floor: 250 words, 500, 750, whatever feels right. You open a blank writing surface and write until you reach it. The words can be anything: a thought about work, a complaint about the weather, or three paragraphs of rough frustration that you delete tomorrow. That is fine. That is the point.
When you reach the floor, the current build marks the day. You get a square on an activity grid. The square records that you showed up and wrote some words.
That is the intended experience. There is no audience, comments, likes, or followers. There is no AI scoring your mood or telling you that your emotional patterns look ‘resilient this week’. There is no leaderboard or streak counter sending a notification when you are about to ‘break the chain’.
The words do not have to be good. They just have to get out.
What WordFloor is not
The clearest way to explain the product is to list what we decided not to build. In 2026, a writing product that excludes AI, social features, and gamification raises eyebrows. People assume you are either naive or hiding something.
Here is what we looked at and walked away from.
AI writing assistance. Every writing tool now seems obliged to offer suggestions, completions, summaries, or ‘insights’. We tested it. We watched it turn private writing into a performance. When a machine is watching, you start writing for the machine. You self-edit before you have finished the sentence. The whole point of WordFloor is to get the raw thought out first, without judgement. AI judgement is still judgement.
Mood tracking and sentiment analysis. This came up often in early conversations. Could we detect emotional patterns? Could we show users their mood over time? We could. We will not. The moment private writing becomes a score, a private ritual becomes a report card. People can start writing to game the metrics, or avoid topics because they do not want to see ‘anxious’ on a dashboard. We would rather not analyse what your writing says about you, and we would rather you did not worry about it either.
Social features. There is no feed, sharing, or way to follow your friends’ writing habits. Private writing is not a public performance. Adding social features would change what it is.
Streak mechanics and shame-based retention. ‘Don’t break the chain.’ ‘Your streak is in danger.’ We have used products like this. They can make missed days feel like failure. WordFloor uses a word floor, not a streak. If you miss a day, the grid shows an empty square. There is no red alert and no guilt email. There is just a blank day. Nothing is owed or needs repairing.
Tags, folders, and organisation systems. Most notes apps depend on organisation features. WordFloor does not ask you to organise your writing. The title field is optional. There are no tags in the first version, no folders, no categories, and no smart filters. You write, reach the floor, and move on. If you need to find something later, there is a basic list. That is enough.
Every one of these ‘no’ decisions made the product harder to build because each removed a feature that could distract from the writing loop. Social features create engagement metrics that can make growth look real. AI features create a ‘wow’ moment that can paper over a weak core experience. Streaks create retention numbers that can look impressive in a pitch deck.
Strip all of that away and you are left with a blank writing surface and a number that counts your words. That number has to be enough. Making it enough is the hard part.
The details nobody will see
Here is what ‘not ready’ actually looks like.
We spent three weeks on the local save cadence. In the current build, WordFloor saves your words to local storage before trying to sync them to the cloud. The public product page describes this as ‘save before sync’. The question was how often to save. Every keystroke would be safest, but it creates unnecessary writes. Every ten seconds is efficient, but you could lose a paragraph if your browser crashes. We are testing three seconds.
We then spent another week testing whether three seconds actually felt right, because ‘felt right’ is not a data point.
We argued about the activity grid colour scale for longer than I would like to admit. The grid uses absolute word-count bands, not bands relative to your floor. A day with 300 words always looks the same, whether your floor is 200 or 1,000.
That decision took a full afternoon. Relative bands would make the grid more motivating, but also more confusing. Two users with the same floor could have different-looking grids even if they wrote the same amount. We chose consistency. The grid should tell the truth about volume, not whether you beat your personal target.
The deletion flow is still being refined. WordFloor is being built with privacy, export, and deletion semantics as core constraints, but those semantics are not finished yet. The intended model is that you can keep or delete the words while the square preserves the record of showing up. The exact behaviour of the payload, metadata, backups, and exports still needs to be specified and tested.
The question is not just how to delete text. It is what happens to the activity record when the text is gone. Does the square stay? Does the word count persist? Can you prove that you wrote something without keeping what you wrote?
We think the answer should be yes. The important distinction is between the content and the effort. We have not finished proving that the deletion behaviour matches that distinction everywhere the data can exist.
The sync logic has its own unsolved edges. What happens if you write offline for three days on your phone, then come back online on your laptop? What if two browser tabs edit the same entry at the same time? What if the cloud and local versions diverge so much that ‘last writer wins’ silently destroys a paragraph you spent 20 minutes writing?
These are normal problems for a local-first product to reckon with. The question is whether you solve them honestly or paper them over with a sync conflict dialogue nobody understands.
Why it is taking so long
The honest answer is that we care about getting it right, and ‘right’ keeps moving.
The less honest answer is that the boring parts are hard and the interesting parts are boring.
Writing to local storage is not interesting. Testing sync conflict resolution across three devices with intermittent connectivity is not interesting. Building a deletion pipeline that separates metadata, payload, audit trail, and tombstone is not interesting.
These are the things that determine whether WordFloor feels trustworthy. Trust matters more than anything else in a product built around private writing.
We could ship now. The core writing loop works in the current build. You can set a floor, write words, see the square, and move on.
But ‘functional’ is not the same as ‘ready for public use’. The sync indicator does not communicate clearly what is happening. The word count update feels laggy on slow connections. The completion state, the moment when you reach the floor, needs to feel calm and clear rather than like a celebration.
Getting that moment right means a few pixels and a few milliseconds of animation. Those details matter more than any feature we could add.
What we have learned about building WordFloor in 2026
Every conversation about WordFloor follows the same pattern. Someone hears the pitch, nods, and asks how we will differentiate it.
The answer is that we are not adding more. We are deliberately not adding things.
While writing tools are adding AI, social features, and gamification, WordFloor is being built without them. That choice may make it disappear. It may also make it useful to people who are tired of every writing tool trying to be their therapist, coach, or social network.
We are betting on the second possibility.
The public product page describes WordFloor as ‘private by default’, with no audience and no AI judgement. That is the promise we are trying to make concrete. It does not mean the storage, sync, retention, or deletion work is finished. Those details still need to be tested and explained before the product opens to the public.
WordFloor is a writing tool, not therapy, diagnosis, or crisis support. It does not assess your mood or respond to emergencies.
What is coming
We are not going to give you a launch date. There will be no ‘Q3 2026’, early-access promise, or countdown timer.
What we will say is that the writing surface is close. The floor mechanic works in the current build. The grid is almost there. Sync, deletion, retention, export, and privacy details need more time. WordFloor is not open for public use yet.
We are building WordFloor for people who want somewhere to put rough, unfinished thoughts without turning them into a performance. People who want a daily writing ritual, not a productivity system. People who are tired of writing tools trying to be their therapist, coach, or social network.
If that sounds like you, you can join the WordFloor waitlist. Joining means you may receive early-access invitations, product updates, and launch information. It does not guarantee access.
Until then, write something. It does not have to be good. It just has to get out.
WordFloor. Private daily writing. No audience. No judgement. Just today’s words.
