My role
Senior product designer: strategy, interaction and visual design, design system, QA
Build
AI-assisted: Claude wrote the code to my specifications
Stack
Vite, plain JavaScript, Web Audio API, SVG


The same kit on a laptop and a tablet: the drawn drums in the Hot Rod look with their keys, and the pad grid in Neon Synthwave.
Overview
PROBLEM
Browser instruments are usually toys or tied to licensed samples. I wanted a realistic kit that plays with multi-touch on a phone and ships as open source.
APPROACH
I wrote the rules before any code, built it in three phases, then shipped six releases shaped by playtests with friends and colleagues.
OUTCOME
v3.1 is live: 3 kits, 8 looks, lessons, layered recording and a concert mode, as a drum kit or a pad grid, by touch, mouse or keys.
kits with their own synthesized sounds
pads: touch, mouse and keyboard
lessons and style beats
audio files, images or outside icons in the app
Process
I used the Double Diamond, but not in a straight line: from v2.0 on, every release went back to people. I watched friends and colleagues play first, then asked what they expected and why, refined, and released again.
Discover
Reference apps, a question round, goals
Define
Brief, principles, phases, information architecture
Develop
Concepts, rules, iterations, playtests
Deliver
Build, release, impact
Small loop: fix and re-test ยท Big loop: back to people every release
Stages 1โ2
Discover and Define
A full product cycle with the AI writing the code to my specifications. Before any code: a question round on scope, sound, input, devices and hosting, and a review of drum apps for interaction patterns (not their art).
Constraints: lightweight and secure, functional first; realistic, not cartoon (true proportions from above, a 14โณ snare and 22โณ kick); every device, by mouse, keys or multi-touch.
Principles: decide the layout early and the look late. Changing the layout means reworking every screen, while colours and textures are stored as reusable values (design tokens) and change with one small edit. A rules file told the AI what was in scope, the quality bar, and when to stop and ask me.
A small product doesnโt need the traditional UX route of full, high-fidelity mockups for every screen, so I kept it lean: concepts first, then functionality, then the UI layer. Quick concepts set the direction, a working build proved the functionality, and the look went on top.
PLAY
Sound on the first tap, keys on a keyboard, thumbs on a phone.
LEARN
Step by step or in time, with only the drums you need lit up.
PERFORM
Record in layers, download the audio, or play to a cheering crowd.
PHASE 1
The kit, sounds, EQ, recording and lessons where the kit plays best.
PHASE 2
Upright phones and tablets brought to the same standard.
PHASE 3
The pad grid, for upright phones and beginners.
Toolbar: primary actions only, in priority order: Rec, Concert, Learn.
โฎ menu: rare settings: EQ, metronome and tempo, theme, keys, flashing lights, guide and reset.
Customize: sound (Kit, EQ) apart from look (Look, Background, Sticks), with the view choice above both, because it decides which apply.
Stage 3
Develop
A grid of 8 pads was quickest, but it does not read as a drum kit (it returned later as an optional view). The real choice was between the two ways people picture a kit:
FRONT VIEW: REJECTED
Iconic, how an audience sees a kit, and instantly recognisable.
But every head is seen at an angle: circles flatten into thin shapes, and even tilted a little, they stay slivers.
Too small for a thumb: scaled to a phone, most heads were only about 22โ28 px tall, half a comfortable target.
Space goes to scenery: stands, legs and empty stage take the screen instead of playable drums.
TOP VIEW: CHOSEN
The drummerโs view: looking down at the heads you actually hit.
Every piece is a separate, round target the shape of its real playing surface, at true relative size.
Fills the screen: wide and shallow, so on a landscape phone every drum is big enough for a thumb.
The trade-off is that it is less iconic at a glance; realistic hardware, chrome, lathe rings and soft shadows win that recognition back.
Invisible, generous hit areas sit under the art, so small cymbals stay easy to tap and the artwork can change without touching input.
Rounds 1 and 2 set the layout: how many instruments there were and how the features would fit. I wasnโt after realism yet, so they looked cartoonish. Round 3, a front view on a concert stage, looked good, but its flat drum heads left more empty space than playable drums. Round 4, top-down, filled the screen with thumb-sized drums and set the direction; the realistic look came later, during the build. The concert idea survived as a feature.
Round 1
Flat shapes: where every instrument and feature goes.
Round 2
Illustrated kit: recognisable, but cartoonish.
Round 3
Front view on a stage: great mood, but flat heads and wasted space.
Round 4
Top-down: every drum fills its space, thumb-sized. The chosen direction.
System
I set the interaction rules and turned key decisions into reusable values (colours, spacing, sizes) and one clear model of how recording behaves. The AI applied them everywhere, and automated checks kept them in place across six releases.
Reach: the drums sit where thumbs rest; rarely used settings live in the โฎ menu, so there are fewer choices up front (Hickโs law); small buttons have 40 px tap areas, so they are easy to hit (Fittsโs law).
Accessible by default: text and edges stay readable in both themes (WCAG 2.2 contrast levels), flashing lights stay within safe limits, and motion calms down for people who turn it off.
Colour means something: saffron means act, red means recording now, and green stays โon timeโ in lessons.
Nothing disappears without a reason: options that do not apply are dimmed, with a way back.
Grouping: pads are grouped by instrument family, one family and one colour per row, so related drums read as a set (Gestalt).
Status always visible: a countdown, a red Rec with the time, one red Stop and a live timeline.
Scope: lyrics and microphone recording were postponed as โcould haveโ (MoSCoW), to keep the kit instant to play.
One colour per instrument family, tuned for each theme so it stays clearly visible on the page (at least 3:1 contrast, WCAG 1.4.11).
I mapped all six recording states before redesigning the controls, with one rule: no button changes meaning under the thumb. Play only plays or pauses, Stop only ends recording, Add layer only starts one.
Every control in every recording state: available (saffron), showing the recording signal (red), or greyed, but never hidden or moved.
One spacing system in two sizes: roomier on desktop (20 / 16 / 12 px for padding, corners and gaps) and tighter on phones (14 / 8 / 10 px), with every control the same 32 px height.
Desktop and upright tablet: labelled toolbar, recording bar above the kit, text tabs and keys shown; Concert in 4 rows on desktop and a full crowd strip on tablets.
Phone, sideways: a one-row icon toolbar, the recording panel as a column beside the kit, icon tabs, keys hidden.
Phone, upright: the kit at the bottom under up to 15 crowd rows, the recording panel edge to edge and floating over the crowd, and Customize stepping aside while recording or learning.
Recording ยท v2.0 to v3.0
From v2.0 on, I handed every release to friends and colleagues, on their own phones and laptops. First I watched them play without any guidance; then I asked what they expected and why, in the spirit of a contextual inquiry. Every hesitation became an annotated screenshot and a fix, checked on 9 screen sizes. Each topic below follows one problem from v2.0 to v3.1.

After
Full-width rows, the player on the right, layer actions under the layers.

The countdown sits in the middle of the screen, where the eyes already are.

Before
Two Stops on screen at once, a REC badge, and a red outline on the panel.

After
One red Stop beside Add layer; the player pauses instead of stopping; marks in each instrumentโs colour.

Before
A recording and a lesson open together.

After
Opening a lesson stops the take, keeps it, and says so.

Before
The note pointed back up to the toolbar.

After
Rec right where the eye ends.
PROBLEM
Layers were a few pixels wide and nothing said when recording began. Later, five cues and two Stop buttons fought for one state, and a recording and a lesson could run at once.
WHAT I DID
Full-width layer rows and a 3-2-1 countdown (v2.0); one red Stop and a real Pause, so no button changes meaning (v2.3); recording, lessons and Concert take turns and the take is kept (v3.0).
IMPACT
Layer rows grew from a thin sliver to the full 800 px width, signals saying โrecordingโ went from 5 to 2, only one mode shows at a time, and a take canโt be lost.
Phones ยท v2.0 to v3.0


On an upright phone: Concert with the drums in thumb reach, and the lesson controls tucked into a โฎ menu.

Dashed outlines show the invisible tap areas (drawn for this picture only).
PROBLEM
Phase 1 was landscape: a two-row toolbar with seven controls, a kit floating above the thumb zone, and small controls too close together to give each one a 44 px tap area.
WHAT I DID
A one-row toolbar cut to four controls, the recording panel beside the kit when sideways, the kit anchored to the bottom under a bigger crowd, and 40 px tap areas that never reach a neighbour.
IMPACT
The toolbar shrank from 104 to 48 px tall, the phone crowd grew from about 174 to 435 people, and no button takes anotherโs tap on any of 32 phone screens.
Pad grid and looks ยท v2.0 to v3.0
Why a second view: the drawn kit is wide, so on an upright phone it fills only a thin strip. The pad grid puts the same drums, keys and sounds into big tiles grouped by family, so beginners get large, clear targets and upright phones get a full-screen instrument, while the realistic kit stays one tap away.
Drag to compare: the same drums, keys and sounds, as a realistic kit or as a grid grouped by family.


The final grid on desktop and on an upright phone, where it sits within thumb reach.
Drag to compare: my review marks on the first grid (names in the corners, every gap the same), then one spacing system and borders in each familyโs colour.

On phones, in the light theme: full-width grid, icon-only Customize tab, no key badges.

One dragon across the whole grid, each tile showing its slice; a band of light sweeps across once per visit.

Neon Synthwave: a sunset and a neon horizon running behind the pads.


The Look tab before and after: tiny grids that all looked alike, then one large sample pad per look.

Cut
Round five of the dragons on the drums: well drawn, but the faces still read as a cat, so I removed them.
PROBLEM
The wide kit fills only a strip of an upright phone. The first grid had cramped names and edges that were barely visible (1.2:1 contrast), and its looks only recoloured the pads, so every pad looked the same.
WHAT I DID
A pad grid view of the same kit (v2.0); one spacing system and borders in each familyโs colour (v2.3); every look drawn across the whole grid, with one large sample pad per Look card (v3.0). Dragon faces that still read as a cat after five rounds were cut.
IMPACT
A full-screen instrument within thumb reach, clearly visible borders (at least 3:1 contrast) in both themes, and eight looks that dress both views.
Learning ยท v2.0 to v3.0

Only the lessonโs drums play; the rest dim and stay silent.

Before
A tiny โGet readyโ, one colour for every drum, big labels over the drawings.

After
The count-in big over the kit; steps with a note tinted in their drumโs colour, the one under the line lit.

After a run, a card offers the next step; the main one sits on the right.
PROBLEM
In playtests, friends hit right on the beat and still saw โMissโ, a result button appeared out of nowhere, and the AIโs idea of playing unused drums quietly would have felt like the phoneโs volume dropping.
WHAT I DID
Hits are now timed from the moment of the tap, unused drums stay silent and say why, the drum to hit glows on the beat, and a result card suggests the next step, with speed levels from 50 to 150% and a short four-step tour.
IMPACT
In a timed test every hit scored within a few milliseconds of the beat, and first-time players get one short tour, then are left alone.
Craft details
Details that make people smile in the first ten seconds: a ripple where the stick lands, a crowd that cheers on crashes, a ba-dum-tss on the first touch, coffee steam and an amp cable. All drawn or synthesized in code, never tappable, and still for reduced motion.

8 looks dress the kit itself: Neon Synthwave on the Spotlight stage.

Dragon on the stage rug, with a spinning record player and a swaying plant.

Classic Studio on Plain: sketched doodles, a music stand and clip-on mics.

Every hit answers back: a ripple where the stick lands, cymbals that wobble, a beater that swings and a hi-hat that lifts.

A crowd that listens: it bobs to the tempo, raises hands on crashes, jumps when you play fast and cheers when you stop.

Concert with the pad grid: laid out under the lights; tiles let the lights glow through.

A steaming coffee, an amp with its cable, a radio.
Testing ยท v2.0 to v2.3
What testing showed in v2
How I tested
Every release went through two rounds. Round 1: my own review on a phone and a laptop. Round 2: friends and colleagues played it on their own devices; I watched first, then asked what they expected and why. The biggest problems from each round are below; smaller fixes are listed at the end.
Round 1 ยท My own review

Before
Seven controls competing for attention.

After
Four controls, in priority order.
Toolbar
What I found: Seven controls competed for attention, and on phones the toolbar wrapped to two rows.
How I fixed it: Four controls in priority order, on one row: Play, Loop and the timer moved next to the layers, and the metronome into the โฎ menu.
Round 2 ยท Playtests with friends and colleagues

Before
Layers were barely visible between the buttons.


After
Full-width layer rows with the player on the right; held sideways, the panel sits beside the kit.
Recording panel on phones
What I found: The panel squeezed the layers into thin slivers between six buttons, and on phones it pushed the drums down the screen.
How I fixed it: Full-width layer rows with the player on the right, on every screen; held sideways, the panel moves beside the kit instead of pushing it down.

Before
Three thin rows of audience, and the drums floating mid-screen.

After
The drums near the bottom, in thumb reach, under a crowd of about 435.
Concert on phones
What I found: The drums floated mid-screen under a thin crowd, above easy thumb reach.
How I fixed it: The kit now sits near the bottom, and the crowd grows row by row to fill the space above: about 435 people instead of 174.
Smaller fixes in v2
Testing ยท v3.0 to v3.1
What testing showed in v3
How I tested
The same two rounds. Round 1: after v3.0 went live, I reviewed it myself. Round 2: friends and colleagues played it on their own phones and laptops. Two bugs only showed up with a finger, never with a mouse, so my automated checks had missed them. Smaller fixes are listed at the end.
Round 1 ยท My own review

Unused drums
What I found: I dimmed the drums a lesson doesnโt use, so its own drums stand out. But the dimmed drums still played when tapped, and the only explanation was the tip at the bottom of the screen.
How I fixed it: They are now truly off while a lesson is open: a tap plays nothing and shows a note right on the drum, โThis lesson only uses Kick and Snareโ.
Round 2 ยท Playtests with friends and colleagues

Lesson bar
What I found: People didnโt notice the lesson bar opening at the top of the screen.
How I fixed it: It now pulses saffron the first time a lesson opens, to catch the eye.

In time or Step by step?
What I found: โIn timeโ and โStep by stepโ meant nothing to first-time players.
How I fixed it: A short four-step tour explains them when a lesson opens, until โDonโt show againโ is ticked.

Before
From a playtest: people missed the tip at the bottom, far from where they were looking.

After
The tip opens in the middle of the kit, then glides to the bottom.
Lesson tip
What I found: Nobody noticed the lesson tip at the bottom of the screen.
How I fixed it: It opens in the middle of the kit first, then glides to the bottom.

Before
The โturn sidewaysโ tip stretched over almost the whole screen.

After
The tip stays closed during a lesson, and everything fits the screen.
Touch-only bugs on phones
What I found: The โturn sidewaysโ tip covered the whole screen and the Customize tab vanished. My automated checks used a mouse, so they never saw these touch-only bugs.
How I fixed it: Both fixed, and the checks now tap like a finger on a phone.
Smaller fixes in v3
Stage 4
Deliver
I owned the what and the why; the AI wrote the how, inside rules I wrote, step-by-step approvals, and automated checks I set.

Playtest notes in, a proposal out: the AI suggested each change and waited for my OK before touching the code.
PUBLIC REPO
CLAUDE.md: the rules for the AI: scope, quality bars, copyright, and โpropose, then wait for my OKโ.
src/: audio, input, kit, learn, ui and styles, one job per file.
docs/: the plan and its decisions, plus a log of every outside asset and its licence.
README: what it is, how to run it, and the folder map.
PRIVATE NOTES, NEVER UPLOADED
A start-here handoff, so every new session picks up with the full context instead of from zero.
The decision log, reversals included, and the open to-do list.
Test scripts and case-study screenshots, ready for the next release.
Kept out of Git, with a search for names, emails, paths and keys before every commit.
RULES I TUNED ALONG THE WAY
Propose, then do: โwait for my OKโ was too broad and slowed small tweaks, so it split in two. Small changes go ahead; installs, deletions, commits and anything public still wait for my OK.
Ask, donโt guess: after the AI once guessed a project address, it now asks before assuming, and I check the setup first.
Decisions on record: a few design details kept flipping, so every piece of feedback and every design decision is logged in the project notes, and each new change is checked against them.
Read first, update last: every session starts by reading the notes and ends by updating them, so no decision lives only in a chat.
v2.0
The full product: kits, recording, lessons, Concert
v2.1
Pad grid sizing and phone polish
v2.2
A fuller crowd on phones
v2.3
Spacing system, one Stop, drums in thumb reach
v3.0
Looks on the pad grid, new lessons, one mode at a time
v3.1
Playtest fixes and names on the drums
Read the full notes for every release on GitHub
Results
releases, v2.0 to v3.1, each with notes
screen sizes checked automatically after every change
automated checks passed before v3.1, so nothing broke
buttons that take anotherโs tap, across 32 phone screens
signals saying โrecordingโ on screen at once
drums that move when Rec is pressed on a phone
looks that dress both the drums and the pad grid
known vulnerabilities at release
A live, open-source product on desktop, tablet and phone: 3 kits, 8 looks, 8 lessons, 10 style beats, layered recording with pause, Concert and a pad grid.
Production quality: 167 source files, about 13,800 lines and only one outside package (the build tool); React, TypeScript and a sampled kit were set aside as too heavy for the job.
Insights
The AI optimises the component; the designer owns the screen, so every review checks its defaults against the whole screen.
My rule โnothing that needs attributionโ was taken literally and nearly removed the build tool; precise rules and true proportions produced original art.
Quiet unused drums read as the volume dropping, so I chose silence with a visible reason.
The AI slipped most on file length (about 200 lines a file). Rewording it to โsplit before it growsโ helped, but a fact belongs in a check, like โevery pad reachableโ, now tested on 9 screen sizes.
Five rounds of dragon faces drawn in code never stopped reading as a cat. Cutting them kept the release honest, and the decorative art that worked stayed.
Drawing and synthesizing everything in code kept the download tiny, made a new look a quick colour change, and removed licensing for sounds and images.
Next time: every device in scope and a live link from day 1, a time-box for exploratory visuals, and moderated tests with first-time players on their own phones.
Result
Drum Kit runs on desktop, tablets and phones, with touch, mouse or keyboard. It is open source under the MIT licence.





