What Particle Refuses

Most software, when it is asked what it is, lists what it does. Particle is easier to describe by what it refuses to do.

This is not a marketing posture. It is the most truthful way we can describe how the product was built. Every meaningful design decision over three years has been a no before it has been a yes. Every feature shipped is the small remainder after a longer list of features rejected. Every screen was drawn after several other screens were unbuilt. The shape of the product is the shape of seven specific refusals — and the refusals are doing more work, by quite a wide margin, than the features.

If you have read An Emma of Your Own, you have already met the principle that explains this. An Emma is a wall, not a cushion. The architecture that lets a creative life survive is not the one that accommodates everything. It is the one that refuses, cleanly and without apology, the small number of things that would otherwise erode the work. Particle is software for that wall. So: the wall.

Below, each of the seven refusals: what we will not build, what tool you have already used that does the opposite, what the research says it costs you, and what we built instead.

#1. No streaks

We will not show you a number that increases for every consecutive day you use Particle, and breaks the day you do not.

The tool you know that does the opposite is Duolingo. Strava does a softer version. Most habit-tracking apps live or die by it. The mechanic is well-engineered: it works. Streaks measurably increase short-term retention because they exploit one of the deepest asymmetries in human cognition — loss aversion. Losing the streak feels worse than gaining it ever felt good, so the user keeps showing up to avoid the loss.¹

The cost is in the second-order behaviour. A streak does not measure whether the work was useful, only whether the user touched the app. Within weeks, users start performing the smallest possible action that preserves the streak. Duolingo users complete a one-question lesson at 11:58 PM in a hotel bar. Strava users go for a 0.6 km "run" in a parking lot. The work has been replaced by the symbol of the work. The streak has eaten the practice it was supposed to support.²

Worse, when life eventually breaks the streak — a flu, a bereavement, a holiday — the user often abandons the practice entirely, because the number was the practice. Streaks are exquisitely tuned to make ordinary human variance feel like failure. The day you needed grace, the streak punished you. Many people never come back.

What we built instead: an accumulating count that only goes up. The number records what was made, never what was missed. Read Why We Count Up for the long argument.

#2. No badges, levels, or scores

We will not give you points. We will not promote you to silver, gold, or platinum tier. We will not show you a "productivity score" out of one hundred.

The tool you know that does the opposite is — well, almost everything. Stack Overflow has rep, Reddit has karma, GitHub has contribution graphs, Notion shows your "streaks and stats", RescueTime gives you a daily score, dozens of habit apps award badges for "100 days of focus". The mechanic, again, works in the short run. Extrinsic markers are easier to perceive than intrinsic progress. They feel like growth.

The cost is the most-replicated finding in motivation psychology of the last forty years. When extrinsic rewards (badges, points, scores) are introduced for an activity that someone already finds intrinsically meaningful, the intrinsic motivation decays. After the rewards are removed, behaviour falls below baseline. This effect, first documented in Deci's 1971 puzzle-solving studies and confirmed across hundreds of studies in the meta-analysis tradition, is so robust that it has its own name: the overjustification effect.³

Translated: introduce a badge, and you teach the brain that the work was being done for the badge. The badge is now the reason. Take the badge away and the reason goes with it. We are not interested in users who do focused work because Particle awards them a badge. We are interested in users who already, alone, decided focused work was worth doing. The product's job is to hold the structure of that decision — not to overwrite it.

What we built instead: nothing. The work is enough. The fact that you sat down and did it is enough. The number that goes up is a record, not a reward.

#3. No "productivity score"

We will not produce a single numeric assessment of your day, your week, or your year.

The tool you know that does the opposite is RescueTime, the older Daily Health Score apps, every "AI productivity coach" that has launched in 2025-26. The mechanic produces what looks like clarity: a number from 0 to 100, an A through F, a coloured ring. The user, allegedly, can act on this.

The cost is Goodhart's drift. Goodhart's Law — when a measure becomes a target, it ceases to be a good measure — has been observed in domains as varied as British inflation policy, school exams, sales commission structures, and self-tracking applications. The user begins, within weeks, to optimise for the score rather than for the underlying state the score was meant to summarise. They take a meeting that bumps the score; they refuse a thinking walk that would have lowered it. The number ate the day. The number did not even know what the day was for.⁴

The more honest answer is that no single number can summarise whether a knowledge-worker's day was meaningful. Whether it was meaningful depends on what the worker was trying to do, what they decided not to do, what they noticed in the gaps, what they refused. The summary that compresses all of that into 73/100 is, structurally, lying. It is a parrot, in Han's sense — producing a confident-sounding output from a system that does not know what the output means.

What we built instead: a record. Hours, sessions, intentions, what you said the day was for, what you did with it. Stored, never summarised, never scored. Available to you to read in any window — the day, the week, the year — in case the shape becomes legible to you. If a shape comes, it comes from your reading, not from our rating.

#4. No leaderboards, no public ranking

We will not show you how you compare to other Particle users.

The tool you know that does the opposite is GitHub's contribution graph, the X profile that displays your follower count, every coding-bootcamp leaderboard, every fitness app's "friends" tab. The mechanic exploits a much older one: social comparison, the human reflex to evaluate the self against the visible performance of others. It is one of the most powerful drivers of short-term engagement available to any platform.

The cost has been documented since the 1950s and re-documented every time a social platform re-discovers it. Upward social comparison (looking at someone doing more) reliably reduces self-rated wellbeing and increases anxiety; downward comparison (looking at someone doing less) reliably reduces effort. Both directions hurt — one through despair, the other through complacency. Where the comparison is made structural and continuous — a ranked board, a public count — the documented effect on wellbeing and intrinsic motivation runs negative, even in cohorts where short-term engagement metrics rise.⁵

There is also a deeper reason. The work that matters is, almost without exception, not legibly comparable. A novelist writing the third draft of a difficult chapter is not doing the same work as another novelist writing the second draft of an easy one. A founder writing a strategy memo is not doing the same work as a founder writing code. Putting them on the same scale lies about both. The comparison is a category error before it is a wellbeing problem.

What we built instead: a private record. Particle has no social graph. There is no friends-list, no follower count, no shared dashboard. Other users do not exist inside the product. Your work is between you and your day.

#5. No push notifications that interrupt the work

We will not interrupt you. We will not ping you. We will not send you a daily reminder. We will not "nudge" you into starting a session.

The tool you know that does the opposite is — most of them. Slack, email, the calendar app, the habit tracker, the meditation app, the productivity coach. The notification mechanic is the engagement engine of modern software, and it is profoundly effective at the metric it serves: time-in-app.

The cost is the single most-replicated finding in attention research of the last twenty years. A single interruption, even one ignored, costs roughly twenty-three minutes before the worker returns to the cognitive depth they had before the interruption. The mechanism is attention residue: the interrupted task leaves a partial trace in working memory that the new task has to compete against, and the partial trace decays slowly. Notifications are not free, even when dismissed. They are not free because they were dismissed.⁶

There is no version of "smart notifications" that solves this. The dismissal itself is the cost. The brain has already left the room.

What we built instead: silence. The Particle window does not interrupt. Sessions begin when you start them. The day ends when you say it ends. If the architecture of your day requires a reminder to start, the reminder belongs to your calendar — a tool whose job is reminders — not to the workspace whose job is not reminders. We refuse to share the calendar's job. The kind of work we want to support cannot be invited; it can only be defended.

#6. No AI coach that lectures

We will not put a generative model between you and your work and have it tell you what you should be doing.

The tool you know that does the opposite is the wave of "AI productivity coaches" that have launched since 2024 — apps that observe your day, generate confident-sounding analyses, and tell you what to focus on. Some are well-meaning. The mechanic, in every case, transfers some portion of the user's agency — the deciding-part — to a generative model.

The cost is the central thesis of The Machine That Cannot Think: a generative model cannot make this decision because the model does not know what your life is for. It can produce confident output. It cannot produce the contemplative middle that a real decision about your day requires. When it nonetheless produces a recommendation, and the user accepts it, the user has displaced their own agency onto a system that does not have the kind of inside the agency requires. The cost is not just bad advice (though usually it is bad advice). The cost is the slow atrophy of the human deciding-muscle that is, after the agent wave, the only muscle left.⁷

What we built instead: the coach that never gives advice. Read The Coach That Never Gives Advice. Particle's intelligence layer asks questions, surfaces patterns, holds memory. It will not tell you what you should be doing. It will help you remember what you said you wanted, and what your last weeks have actually looked like, and let the deciding stay where it has to stay — inside you.

#7. No options sprawl

We will not give you a preferences pane the size of an aircraft cockpit.

The tool you know that does the opposite is, again, almost everything in the Notion / Obsidian / Linear / configurable-IDE family. The mechanic is well-intentioned: respect users by letting them make every choice themselves. The honest reading is also less generous: every option is a deferred decision, an admission that the team building the product could not commit to a default. Every preference is a question handed back to the user.

The cost is the option tax. Every additional choice is an additional decision, and decisions compound: each one measurably lowers the quality and willingness of the decisions that follow it. A working day's decision budget is finite, and it is, in any honest accounting, the most finite thing about a knowledge-worker's day. A piece of software with three hundred preferences spends the user's most precious resource on its own configuration before the user has made a single decision about their actual work. Most users, faced with this, accept the defaults — and now the software has a setting screen the size of a room that no one will ever use.⁸

There is a deeper cost. The product that lets the user customise everything has, structurally, refused to have a position. It cannot be a wall, because it has been built to bend. It cannot be Emma, because Emma is a person with views. We want Particle to have views. We want it to be opinionated about how a working life should be held. A user who disagrees with our views can use a different tool — that is fair. A user who agrees with our views deserves a tool that acts on those views by default, without asking them to ratify each one in a settings pane.

What we built instead: defaults. Strong defaults. Defaults the team has argued about and re-argued and now ships without options. The number of preferences in Particle is small on purpose. Each one was earned. Each one represents a decision the team genuinely could not commit to, after argument, and so chose to honour the user's variance. Most decisions, we did commit to. The product is the commitment.

#What this leaves

These seven refusals are why Particle, opened for the first time, looks small. They are why the screen is mostly blank. They are why there is no "tour", no "dashboard", no "achievements" tab. They are why the preferences pane fits on one screen. They are the reason a user who has been trained for fifteen years on dopamine-engineered productivity software opens Particle and feels, briefly, that something is missing.

Something is missing. The missing parts are the ones doing the work.

What is left, after the seven walls are in place, is a small, quiet space. A timer that counts up. A document. A few keystrokes. A record of intentions, kept without judgment. Sessions that begin when you start them and end when you say. A day that closes when you close it. The Particle Loop — Capture, Plan, Execute, Complete, Reflect, Align — held by the structure, not by the streak.

That is the whole product. It is also, we think, the only software in the category that is honestly trying to serve the four-and-a-half hours of a working life that have ever produced anything that lasts. The other tools are serving a different layer — the throughput layer, the visible-output layer, the layer the agents are now eating. We chose the layer underneath. We chose it on purpose. The walls are how we keep it.

A good Emma says no. We try to.

#References

For the design pieces this article extends, read Why We'll Never Add Streaks, Why We Count Up, The Coach That Never Gives Advice, and The Case Against Notifications. For the philosophical anchor, read The Machine That Cannot Think and The Freedom That Enslaves. For the principle behind every refusal, read An Emma of Your Own and The Emma Problem.