- 67 onboarding screens, cut to 23 for the first release
- You make an alarm by pasting a TikTok, Instagram or YouTube link
- Icon set, coin economy, brand and the launch films
- Shipping for the RevenueCat Shipaton
- Product UI and UX, and the brand it sits in
- It shows its working before the answer, so a non‑engineer can judge it
- Designed and built the website and the interactive demos
- Positioning, and securing the first paid enterprise customer
- Retention and growth, 36 screens
- Drawn against the live app, every claim checked and dated
- I chose the two problems to work on
- Every screen badged live, designed but not shipped, or mine
- Raised $2 million, including Lion/Kirin's first investment in an external startup
- One system across retail, online, venues, events and merch
- More than 500 retail venues, around 400,000 litres a year
- Brand, packaging, ecommerce and the store screens
Wakebait is an alarm clock where the alarm is a meme. Pick one from the stock library or paste a TikTok, Instagram or YouTube link, and tomorrow morning that’s what wakes you up.
- Design
- Every screen here is mine
- Build
- In progress, Pete Nicholls leading engineering
- Shipped
- Nothing yet
- Scope
- 23 of 67 onboarding screens, plus a 13 screen app
Apple opened AlarmKit at WWDC25, and anytime Apple opens something up it presents as an opportunity. An alarm is retentive by nature: it goes off whether anyone opens the app or not. Memes are viral by nature, and the content already sits where the audience is. Retention on one side, distribution on the other.
It’s a silly idea. I’ve seen too many silly ideas work to not try one.
Onboarding
Ten screens where nothing goes wrong, and thirteen for when it does.
It says what it’s, and stops
It’s an alarm clock. The alarm is a meme. That’s the app. Then a button reading Go on then. No feature list, because the premise either lands or it doesn’t.
It shows the core of the app before it asks for anything
Three slides, no input. A clip on Instagram: That sound? That’s an alarm. The share sheet: Send it to Wakebait. Two taps. It’s in your alarms. Then tomorrow’s card: Tomorrow, this bloke wakes you up. The core of the app in the first screens, before a single question. It’s not in the MVP because marketing can do the same job, and it’s one less thing to build.
It asks what kind of noise
What should wake you? Classic, the ones everybody knows. Cooked, nobody involved is okay. Rough, you will be awake. It costs nothing to answer and makes the next screen personal.
It shows you tomorrow
One card: the sound name in full height type, a snooze circle, a stop button. You know what happens at 7:07 before anyone mentions a permission or a price, so every ask after it’s small.
Two ways in, and the second one is the product
Stock meme, or your own? The grid is the easy path and the link is the point: TikTok, Instagram, YouTube. Links break constantly, so the failure is designed: That one’s locked down. Try another.
The screen that decides whether the app lives
An alarm that can’t notify isn’t an alarm, and iOS only asks once, so the app asks first: Nothing rings without this. People still say no, so the declined path is a real screen in the same tone. Declining isn’t a mistake, and dressing it as an error says otherwise.
It hands you over
The glow stops, the background goes flat black, and you’re in the app with one alarm for tomorrow. Onboarding was selling. This is a tool.
The app
Twenty six screens designed, thirteen being built. Five areas, and what each one lost.
Home
Home is a dashboard for what’s coming: tomorrow’s alarm on a card, and under it a wake-up streak, a snooze bank, and the alarms your mates had set for you.
The streak is there for the reason a streak is always there. Snooze isn’t banned. It’s banked. Five wake-ups earn a credit. Three systems on one screen, none in the build.
Library
Every sound you have, filtered and sorted, each row carrying its source, because a sound from a mate and one off TikTok aren’t the same object. The first build has no grid and no Mates filter.
Alarms
A list and an editor. As designed it also had Roulette: the alarm fires anywhere inside a window, not on the minute, so you don’t get to know. Some people don’t want to choose, they want to laugh. Designed, not being built.
Add a sound
Four ways in as designed, one in the build. The paste field carries the product twice and they’re not the same problem: in onboarding it’s the whole screen, in the app it’s a card above a library that already has things in it.
Mates
You pick the time, your mate picks the noise, and they don’t need the app to record it. Their sound stays sealed until it fires, because hearing it early would spend it. Same bet as Roulette: the morning pays it off. Designed, not being built.
MVP
Five whole systems went for one reason. We’re shipping the smallest version that can prove the idea, because there are too many unknowns to know whether any of them work, and the deadline is real.
One thing I got wrong rather than descoped. The name screen I built and reverted inside an hour: collected and never shown, it’s a field to fill in for nothing.
Mates went for the same reason. The whole bet is one thing, any meme linked to your alarm, and nothing gets built on that until it’s proven. It comes back first though: your mate records in an App Clip, so whoever records never installs. Every other cut is a feature. That one’s a channel.
How it looks, and why
Two typefaces, one colour, glass, a sky, and a mark.
Two typefaces, two jobs
Onboarding uses Plus Jakarta Sans because it’s still selling, and because I liked it. The app is flat black and SF Pro because at 7am it’s a tool, and it should feel like the phone’s own.
One colour, and it means something
Amber means on. A lit toggle, a button you can press, a mate’s badge. One amber thing per row. Because it means on, it’s never a background inside the app: a big flat block of amber looks like something waiting to be tapped. Onboarding gets to be warm because its warmth sits far below anything you can tap, so nothing there looks lit.
Glass
Every card in onboarding is a thin dark glass over the sky, never a solid. Things set into a card go a step darker, then the controls, then the chips, and each step has one reason. I tried Apple’s frosted material first and it washed most of the colour out of the sky behind it, so the cards are plain black at a third and nothing more. Light-and-opaque is a card, dark-and-opaque is a hole.
The sky
Clouds in the early hours of the morning, an ember under them and a grain over both. It’s polish, and it lives only in onboarding.
The mark
The alarm time is the logo: 7:07 upside down reads LOL. The icon is the mark itself, each digit turned so that it spells it, and on the welcome screen the wordmark flickers on like a neon sign.
How it moves
Tapping a row holds for a beat, because instant advance swallows the tap. Selection is a light impact, finishing is a success.
Sound follows the same rule: out, a beat of nothing, then in. Never a crossfade, because these are people talking and two at once isn’t a blend. The gap is seventy milliseconds and it’s the part that does the work.
Compiler answers plain-English questions about how a product works, and every answer is grounded in your source code. It’s for the teams who support, onboard, sell and operate the product, so they stop filing tickets and waiting on engineering.
- Role
- Co-founder, with Pete Nicholls
- Design
- Product UI, UX, brand, website
- Build
- Pete leads engineering
- Shipped
- Live, in use, paid for
- First customer
- The Lookout Way, healthtech, about 100 people
The knowledge exists. It’s just locked in the code. An AI can be wrong about it, so the interface shows its working before it shows its answer, and says plainly when it’s not sure.
Non-technical users can judge an answer for themselves.
Product
What the first screen asks for, what order an answer arrives in, and who owns each decision.
The first screen asks for nothing
The people who open Compiler work in support, sales, training and operations. They’re not stuck because the codebase is hard to read. They’re stuck because I can’t answer this without filing a ticket and waiting three days. So the first screen asks nothing of them. No repository to choose, no syntax, nothing to set up. What can I help with?, one field reading Ask anything, and three questions anyone could have typed: What changes have been made over the last week? What bugs were fixed recently? Walk me through one of the key features.
The three examples are the entire tutorial. They set the range by showing it, which is cheaper than explaining it and it survives people who never read anything.
Compiler is changing our world. For the first time ever, we’re getting ticket velocity up. Kate Parkin, Chief Customer Officer, The Lookout Way.
The answer shows its working
The workspace in these screens is our own, connected to Fizzy, Basecamp’s open-source project tool, so the questions here are about boards and cards. The repository is public, so nothing on this page is a customer’s.
Ask What can someone see on a public board without signing in? and the answer doesn’t arrive first. The working does. Compiler says in plain words what it’s about to check, marks each step Exploring and then Done, and only when it has a clear picture does it answer, under a heading you could have written yourself. Under every answer sits Compiler can make mistakes. Please double-check responses.
The order is the design. Working, then answer, then the caveat, so someone who can’t read code can still judge what they were told.
Sharing stops at the company
An answer is worth more once somebody else can read it, so any conversation can become a link. The dialog says what the link is before you make one: Create a link to share this conversation with your team. Only members of your organization can view. That second sentence is doing the work of a permissions screen.
The boring screens
Settings is where a CTO decides, so it’s built to be obvious. One panel, a search field at the top, and three groups named for who owns the decision: Account, Organization, Projects. Members have roles, and an invitation is a link that works once and expires after a day, because a link that lives forever is a door left open. The audit log lists every message and every conversation with a link to open each one, so an admin can read what people are actually asking. That’s the point of it. The company gets to see how the product is being used, not a list of who signed in. Repositories sit under the project rather than the organisation, so the permission lives next to the code it covers. It’s not fancy. It’s logic, and none of it’s trying to reinvent the wheel.
More tools, still scoped to the project
A project can be given tools beyond the code it reads. A connector is a remote MCP server: a name, an https URL, and either no authentication or a bearer token. The name isn’t decoration. It prefixes every tool name from this server, so an answer can never hide which system it reached for. The empty state says what the thing is for before anyone has one: No connectors yet. Connect a remote MCP server to give the assistant extra tools in this project. And it sits under the project, beside the repositories, for the same reason they do.
Website
One page, two readers, and the objection answered before it’s raised.
It says it twice
The headline is for the customer success lead: Turn your codebase into something the whole team can understand. The small line above it’s for their CTO: Read-only GitHub access · Connect your repo. One screen, two people, and the security question is answered before the second one thinks to ask it. That line is smaller than the headline on purpose. Visible, not dominant.
With, without
Engineering shouldn’t be the bottleneck. Two cards, five rows each, and the same five labels in the same order on both: access, answers, context, onboarding, productivity. With: your whole team can understand how the product works. Without: understanding depends on engineering. The labels are what make it readable. You land on the same subject twice and the difference is the only thing left to notice. The colour does the rest: the with card is the one that carries the blue, and the without card has none in it.
The question that decides it
The third section is Controlled access, headed Read-only by design. Four cards, one sentence each: choose the approved repo, keep permissions scoped, ask product questions, get grounded answers. The second is the one that matters: Compiler can read authorised code, but can’t commit, merge, deploy or change how your product works. Under it, a CTO who isn’t ours:
Compiler has been added to our list of approved AI tools for responsible use under our Information Security Policy. Tate Johnson, Chief Technology Officer, The Lookout Way.
Security gets a numbered section rather than a line in the footer, because to the person who signs it off it’s not a detail at the end. It’s the question.
Six quotes, and a job on each one
The fourth section is What Compiler unlocks across your team. Six quotes, and above each one the job it came out of: customer operations, sales preparation, training material, system validation, customer success, support tickets. All six are from The Lookout Way. The label is doing the work. Someone scanning for themselves finds the job first and the quote second, so the section answers whether this is for people like them without a page of personas.
The questions people type
The FAQ isn’t written for us. Every question is one somebody types into a search bar: What’s Compiler? Who’s Compiler for? Where do Compiler’s answers come from? Does Compiler store our code? Seven questions, each one a thing a buyer asks before they ask us.
Brand
The mark, the three typefaces, the one blue, and where none of it goes.
The mark is two letters and a full stop
The mark is Co. drawn as one shape: a C wrapping an o, with a square full stop tucked into the C’s mouth. The wordmark uses it as a letter and finishes the word, so the thing at the top of the site reads Co.mpiler. The full stop is deliberate: Co. is how a company writes itself short, and it sits inside the product’s name. It’s the name, not a symbol standing next to the name, and it never stands alone: on the site it’s a letter in the wordmark, and the app carries no logo at all.
Three faces, and each has one job
Geist carries anything you read as a sentence. JetBrains Mono carries the furniture: the eyebrow, the buttons, the row labels, the section headings and the footer. Handjet, a square pixel face, appears four times on the whole page, on the section numbers 01 to 04. Each face has one job and never takes another. The chat mock inside the hero is set in a plain interface sans instead of either of them, so it reads as the product rather than as the page.
Where the colour stops
There’s one accent colour on the whole site. It marks the line that answers the security question, and it fills the card for the way things could be. The card for the way things are now has no colour in it at all.
None of that’s in the product. The app doesn’t have a typeface of its own, it uses the system one, and in fourteen screens colour appears twice: a green dot while a step is running, and a blue link in the audit log. That’s on purpose. The brand does its work in the marketing and at the front door, and this is B2B, so once someone is inside, it needs to be stripped back to the core so it can do its job. No one needs to be distracted.
Purpose is an AI mentor you talk to. It asks well, and the best thing in it’s the conversation. The conversation just never leaves the room. It can’t see the report the same product wrote, and nobody outside your phone ever hears it.
- Product
- Purpose, Mark Manson’s AI mentor
- Brief
- A take-home. I chose what to work on
- Design
- Two flows, 36 screens, all mine
- Drawn against
- The live app, 7 to 10 August 2026
- Shipped
- Nothing. Every screen is a proposal
Retention closes the loop the product already opened. Growth opens one it never could. Two flows, both drawn against the running app rather than the design file.
Every screen is badged live, designed but not shipped, or mine.
Retention
What the record already holds, the seam between the two halves of the product, and the one new object I’d add.
93 words, four ticks and no source
Journey is the record: reports and conversations, tagged and dated, all of it looking backward. The report itself is 93 words and four ticks, and nothing says what a tick means or where any of it came from. Under it sits 258 words of real advice, with named protocols, times, durations and metrics. Anchor Hour, fifteen minutes, weekday mornings. All of it’s set as undifferentiated prose, so nothing in the product can pick it up and act on it.
Chat can’t see the report it just wrote
Ask the assistant what to do next and it says that’s up to you. Push it on Anchor Hour, its own protocol from its own report, and it says I’m not familiar with that. It still recalls earlier conversation, so this is scoping, not amnesia. By morning the dead end has filed itself into Journey as a timeline entry with a pull quote. That’s the one surface genuinely alive here. But entries record. Nothing in that list ever comes back and asks.
Great at asking, silent after
They have already designed the check-in. It exists only in the 2.0 file, on a roadmap marked outdated, and whatever she answers the reply is the same: GREAT JOB, thanks for checking in. The plan doesn’t change, Journey doesn’t change, and nothing comes back. The input is solved. The output is the hole.
Same words, then a choice she can act on
The proposal cuts no copy. The same 258 words, left aligned to match their own section screens, with the named practices weighted and a cadence line under each protocol. Only the ending changes. Today the report stops and hands her to chat carrying nothing. Here it ends in a question, Which one would you actually begin this week?, with the three protocols under it as rows she can tap: Anchor Hour, The One Bet, Weather-Tuned Work. Tapping one is what starts everything after this. A fourth row takes none of them, because a decline is information, not a failure, and it sends Purpose back to question its own read rather than her commitment.
One new object, pinned to the surface that ships
Purpose overrides its own advice downward, which is the move a streak counter can never make, and the rhythm is agreed in the conversation rather than set in settings. What that produces is a Focus: four fields, one of them private to Purpose, sourced from the report’s own insight rather than the quest catalogue. It pins to Journey, in their own card component, so Focus joins Conversation and Report as an entry type. Nothing else about the record changes. While it’s live it’s the first thing Journey shows, and done or dropped it retires into the timeline as history.
The answer changes the plan
Their check-in sheet, untouched, with one change: the question. How did Anchor Hour go, zero to five, five days and not seven because the plan is weekday mornings. The number arrives in the conversation, so she never retypes it and only types the reason. That handoff carrying data is the exact seam the live app is missing. Fifteen minutes becomes five, re-sourced from her own words, and Purpose commits back: I’ll ask you Friday.
Growth
The one claim the product can’t back, the people who can check it, and what the OS does that the app can’t.
The report makes a claim it never checked
The report tells you how the people around you see you. Onboarding never asks any of them. There’s a screen in their own file that would have, sitting between the questionnaires and Analyzing your answers, and it was cut. I asked the app where the claim came from on 8 August. It said I don’t have a source for that, that’s me overstepping.
Nothing in the product can go and check it. Growth today is two cards behind the profile icon: Share with a Friend and Give the Gift. Both are transactions. Send someone a link, or buy someone a membership. Nobody who taps either ends up inside the experience. And the first one doesn’t send a link. Checked live on 9 August, it sends a plain string of text. iOS shows its plain text icon, and there’s no preview, because there’s nothing to preview. A blob of text can never become a card anyone taps. So the claim it can’t back is the one worth going out and checking, and checking it means reaching three people who have never heard of Purpose.
The invitation goes where the blindspot is
Every report has the same four sections, and Blindspots is the only one self-report can’t answer. So the ask sits inside it, in their own Next Insight card, which costs no new component. The sheet it opens is their pattern 472, the illustration-led one for a single decision, not 476, the two-option task sheet, which is where the wall of text came from. It leads with what the ask costs the friend, not with what you get. Nothing to install kills the objection before the words App Clip appear. The referral card in the settings is rewired to the same ask, so there are two doors in, and Give the Gift is left alone because it’s a real product rather than the broken part.
Three people, and the phone does the addressing
Purpose is minutes old here and knows nobody, so the screen composes rather than picks. There’s no list to pick from: no contact picker exists anywhere in their file. Relationship routes the questions, duration filters them, and three is the number, because one gives no cover and five is a lot to ask. It never takes a phone number, so Purpose can’t address the message and the OS has to. Purpose drafts it, she approves it, and she can cut a line. A message that sounds like her gets answered, and anyone who edits the ask has committed to it.
A card, and four questions in chat
What lands on the friend’s phone is Apple’s card, built to Apple’s spec: a 3:2 image, a title capped at eighteen characters, a subtitle at forty three, an Open verb and a permissions line. Four questions is fourteen characters and Two minutes, unattributed is twenty five. Both fit. Open starts an App Clip, so she answers without installing anything. I drew that as four questions and a Next button first, and it was wrong. Chat is the product, so a form shows a prospect everything except Purpose. She presses send herself, because the anonymity promise is only a claim until she acts on it. Then she’s asked once, and the offer is her own report with the answers redacted. The empty state does the selling.
One read confirmed, one corrected
Two days, a floor of two, and no names while she waits, because one reply identifies its author. One push brings her back, and it says changed, not checked, because checked would mean the section is as she left it. What she finds is one card with rules rather than three cards: one view from three people, never attributed. Then the section re-renders. The report had said she was busy. Not one of the three said busy, and two said bored. It’s the first time the product can be told it was wrong.
How it’s checked
Three badges, a date under most claims, and the one thing I couldn’t source.
Three badges and a date
Every screen on both boards carries a status badge and a second tag, naming either what’s broken or which of their components it reuses. LIVE · VERIFIED, then DESIGNED, NOT SHIPPED, then PROPOSED. Under most of them is a line beginning CHECKED, carrying the date I looked, the screen I looked at, or the source. It means you can disagree with the proposal without having to take the diagnosis on trust.
I drew the app, not the file
Where the 2.0 file and the app disagree, I drew the app. The file has a referral card saying friends together break patterns, the live one says break bad habits, so that’s what the board says. The exception is their check-in sheets, which exist only in the file. I used them anyway, badged as theirs, because they’re the team’s own designed intent and because Mark Manson had told TechRadar that check-ins were shipping by New Year’s. Not using them would have meant inventing an answer they had already drawn.
The one thing I couldn’t source
One note doesn’t say CHECKED. It says INFERRED, on the grey tint behind the message bubble, which is my read of how Messages frames a rich preview rather than anything Apple documents. It’s the smallest detail on either board. It gets its own badge because the moment one unmarked guess is in there, every checked line beside it’s worth less.
Where I’d start
Two bets rather than finished specs, in the order I’d build them, and the four things I’d check first.
Retention first, because it’s wiring
Retention leads the build because it’s the smaller lift. Their check-in sheets are already drawn, and the surface the focus pins to is the one that ships today. Almost none of it’s new. The one genuinely new object is the Focus.
Growth second, and four things to check first
Growth is second to build because the App Clip is the heavy lift. Before either, four things: map the hot path with real usage data, work out who’s actually paying, audit the rest of the app including Quests, and re-check the roadmap with the team. Both boards assume answers to all four. None of it blocks the wiring. It runs alongside, correcting where I guessed wrong.
I co-founded an independent beer company and designed everything it touched.
- Role
- Co-founder, 2019 to 2023
- Raised
- $2 million, Lion/Kirin’s first outside investment
- Design
- Brand, packaging, ecommerce and store screens
- Reach
- 500 retail, 100 on-premise, 400,000 litres a year
- Ended
- A share buy-back in 2023, brand discontinued
Every other project on this site is one I could change after it shipped. This one went to a printer. What I produced wasn’t artwork, it was instructions, applied for four years by people I never met, onto surfaces nobody can patch.
The phone beside this is the brand’s social feed, rebuilt down to the chrome. The work in it’s real. The captions and the counts aren’t.
The ninety-five per cent
Craft was five per cent of the market and having the whole conversation. Everyone else was the brief.
Five per cent of the volume, almost all of the voice
Since craft arrived here, beer had been an argument about who owned what and which technique mattered, conducted almost entirely by the people winning it. Craft was around five per cent of the Australian market and held nearly all of the noise: Young Henrys was brewing roughly three million litres a year to fifty-eight thousand followers while Carlton Draught did a hundred and twenty million to four thousand. The people buying the other ninety-five per cent weren’t in that argument and nobody was talking to them without it. The strategy called them the silent majority, and the positioning was four words long. Great beer. No nonsense. Neither screen here has a can in it, and that’s the first thing the positioning cost. A brand whose whole argument is against shop talk has to be willing to turn up without its own product in the frame, or it’s just another beer ad claiming not to be one.
Fifteen seconds on the day, one frame on the beer
Nine-to-five was a social campaign, and it spends its entire length on a truck driver’s shift: the cab, the chains, the highway, the rain. The mark composites in over a schooner at the end, for about a frame. It’s advertising rather than a design artefact, and it’s worth saying so, because that’s exactly what makes it the strongest evidence on the board that the positioning was an edit decision and not a page in a deck. A brand that believed its own line about shop talk had to be willing to come last in its own campaign. The stills were shot to the same rule: five frames of a bloke in his shed, and the beer in two of them.
Reward Yourself is a transaction, not a boast
Reward Yourself has a condition attached, and the condition is the other half of the brand’s own purpose, work hard, play hard. That’s why it could be printed over and over down the side of a carton without going hollow. It’s not the beer claiming to be special, it’s the drinker being told they have earned one. A craft brand would have explained the hops here. This one states a trade and stops, on a flat blue field with nothing else in the frame to read.
Fixed so it couldn’t be got wrong
A mark that could be owned, one blue with no tints of it, and a manual written by the person who then had to follow it.
The tilt is the mark
A logotype would have secured nothing. Anyone can call a business Steel City, beer or not, and set those same two words in whatever face they like. So the equity went into a shape instead: a ring, read as an ouroboros, for renewal rather than heritage, drawn in three dimensions on purpose. Picture a plain band on a table with text cut inside and out, then tilt it until both readings are visible at once. That tilt is the mark. Watch it rotate and you can see the thing being protected: flattened, it stops being distinct and goes back to being a description.
No tints, which is stricter than it sounds
Pantone 2728C, #1444b9, and the guide says in as many words that tones of it aren’t permitted. That single rule does more work than it looks like it does. One unmixed blue survives a fridge door, a tap badge read at five metres and a phone screen at arm’s length, where a family of blues reads as three different companies sharing a logo. It also removes a decision from everyone downstream: there’s no lighter one for small sizes and no darker one for print, so nobody has to choose and nobody can choose wrong.
I wrote the manual and then had to be the client
The foundational documents are copyright-lined to Y0KKE, the Sydney studio I co-founded, and the 2022 campaign work to Steel City Holdings. That’s the arrangement in two lines: the brand was built by my own studio and then brought in-house as the company could carry it, and no external agency was briefed at any point in four years. It also meant I wrote the client manual and then spent three years as the client. It names the typeface, Acumin, chosen because it’s a modern take on an industrial style rather than a face of its year. It fixes clearspace at half the height of the logo, and says the ring may never print below five millimetres square. Rules that specific aren’t pedantry when the person applying them at six o’clock in a venue isn’t you.
Where it went without me
Cropped, blurred, incidental and small: four years of other people applying it, and what survived.
Cropped by whoever was holding the camera
The banner behind the band is cut off by the top of the frame. Two thirds of the lockup is in shot, the rest is gone, and nobody involved in taking that photograph was thinking about clearspace. It still reads, because a ring with a hole in the middle is still a ring when you take a third of it away, and a word isn’t still a word. This is the first of three photographs nobody art-directed.
Recognised with the letters gone
Two schooners being clinked in a pub, and the only branding in the frame is a coaster on the table. Zoom in and the lettering inside the oval is blurred past reading. The mark is still identifiable, from the silhouette and the blue and nothing else. That’s the entire case for putting the equity in a shape rather than a word, made by a photograph taken by someone who had never seen the guidelines.
Small, and in someone else’s day
Five cards in that carousel and the can appears in two of them, filed next to a die grinder and held against a chest in a cluttered shed. It’s small in frame both times and it’s not the subject either time. That was the instruction: the brand is a guest in someone else’s day, not the reason for it. The can on the concrete is the same argument with the shed taken away: still at ankle height, still incidental, still legible.
It launched on tap in April 2021 and in cans the month after. In its last year it moved around four hundred thousand litres through more than five hundred retail accounts and a hundred on-premise venues. Then the shares were bought back and the brand was discontinued. There’s no shelf to point at now, which is the honest reason this section is a phone and not a photograph of a fridge. Every other project here I could fix on a Tuesday.