Every idea starts as noise.
Most ideas for music tools are guesses. A few are brilliant. Most miss. Before anyone writes a line of code, we find out which is which, together with the people who'll actually use them.
- A hunch
I think DJs want this.
Most decisions in tech get made right here. - One story
A DJ once told me…
Still mostly noise. One voice is a starting point, not proof. - A pattern
Several people mentioned the same thing.
Now we are listening properly. - Many conversations
Almost everyone we spoke to described it the same way.
The signal is getting clear. - Real behaviour
We can see it in what people actually do.
Now we are watching, not just asking. - Tested
We tried it for real, and the result holds.
Where we want every decision to be.
Problem, or an answer in disguise?
It's the first trap, and everyone falls in. An idea for a feature turns up dressed as a problem. How you describe a problem decides which answers you can even see. Try it yourself.
“We need an app for DJs.”
An answer. It names something to build and says nothing about what's actually hard.
“Ten minutes before a set, DJs can't find the track they know they own.”
A problem. A real moment, a real person, a real cost, and plenty of possible answers.
“We should add AI recommendations.”
An answer. A technology looking for a reason to exist.
“Between events, promoters lose touch with their regulars.”
A problem. It describes what's going wrong, and leaves the answer open.
How an idea moves through Labs.
Every idea takes the same route, and artists are part of every step.
01Listen
It starts with a conversation about the last time something went wrong.
What happens
We ask about real moments, not wishes. “Tell me about the last time…” teaches us far more than “Would you use…?”. You do most of the talking. When something clearly frustrates you, we dig deeper, and we keep asking why until we reach the cause behind the symptom.
What you do
Tell us how you actually work, including the messy parts nobody posts about.
What comes out
The problem in your words: “When I…, I want to…, so I can…”. Plus an honest list of what we're still only guessing, sorted by how much it matters and how little we know.
How we keep ourselves honest
Can we point to something a person said or did? If not, it's still a guess, and we label it that way.
02Build small
We take the riskiest guess and make the smallest thing that can test it.
What happens
Each guess becomes a clear belief: “We believe… because… We'll know we're right if…”. Then we pick the cheapest way to find out. Often that's a sketch, a clickable mock-up, a short video or a rough working version. AI helps us sketch fast, so we can try several very different ideas before we fall for one.
What you do
React early, while changes still cost almost nothing.
What comes out
A few rough versions, each built to answer one question.
How we keep ourselves honest
Is this the simplest thing that tests the idea? If we could learn it faster or cheaper, we do that instead.
03Test with you
You try it in your own setup, on your own music.
What happens
We watch what you do, not only what you say. We pay most attention to the first few minutes and the confusing moments, because that's where tools lose people. We check simple things. Does it fit what you're trying to do? Is the next step obvious? Does it behave the way you expect? Can you undo a mistake? Can you make it your own?
What you do
Use it the way you normally work, and tell us where it gets in your way.
What comes out
Evidence: what worked, what didn't, and what surprised everyone.
How we keep ourselves honest
Before every test, we write down what we expect to happen. If we already know the answer, the test is biased, so we change the test.
04Decide
Every test ends with a decision, and you hear what it was.
What happens
We write it down in four lines: we believed…, we saw…, we learned…, so we will…. Then we pick one of four moves: build it, change course, park it for later, or stop.
What you do
See what changed because of your input, and tell us if we got it wrong.
What comes out
A decision we can explain, and the next riskiest guess to test.
How we keep ourselves honest
Perfect evidence doesn't exist. We aim for enough to stand behind the decision, in steps small enough to undo.
Five questions every idea has to answer.
An idea can pass one and fail the next. We always test the one we're least sure about first.
-
Valuable
Does solving the problem or creating the opportunity deliver real benefit or impact to users?
-
Desirable
Do users actually want it? Will they choose it? Is solving the problem relevant enough for them?
-
Viable
Does it work for the business? Does it generate valuable outcomes, benefits or opportunities for the company? Can we sustain it?
-
Usable
Can users solve their problem or seize the opportunity with the product and figure out how to use it effectively (without continuous help)?
-
Feasible
Can we actually solve the problem? Do we have the skills, capabilities, and knowledge? Can we build and deliver something?
Does it matter?
Would solving this make a real difference to your work, or to your night?
We compare before and after: preparation time, stress, moments you missed.
Would you choose it?
Out of everything you could use, would you actually pick this?
We show a short explainer or a rough version and see who asks to try it.
Can it last?
Can it keep going in a way that is fair for you and for us?
We talk about price and value early, long before anything launches.
Is it easy?
Can you figure it out on your own, the first time?
First-time tests without any explanation. If we have to explain it, it isn't ready.
Can it be built well?
Will it hold up on real music, real libraries and real gear?
A rough working version on real audio, before we promise anything.
Wanted and worthwhile aren't the same thing. Endless scrolling is wanted. Stretching before a long set is worthwhile. We look for both.
Cheap evidence first.
The first honest conversation teaches more than the fifth month of building. Every extra step costs more and teaches less, so we only climb as far as the decision needs.
A conversation
A few hours to set up. Tells us whether the problem is real.
A sketch
An afternoon. Tells us whether the idea makes sense to you.
A clickable mock-up
A day or two. Tells us whether you'd know what to do next.
A rough working version
A few weeks. Tells us whether it works on real music.
The finished product
Many months. Tells us whether it sells. A very expensive way to learn.
Fast, cheap or certain: early on, we choose fast. There's always time for a more certain test later.
What would you decide?
Real results are messy. Here's a made-up example of the kind of call we make all the time.
We showed a two-minute explainer of a new feature to 40 DJs in the community. 12 asked to try it. We'd hoped for 10.
But in follow-up chats, 3 out of 5 couldn't say what the feature actually does.
And a big software company has just announced something similar.
Pick a move to see how we'd think it through.
Build it
Tempting. But if people can't explain it, they probably wanted something a little different from what we'd build. Building now means guessing which version.
Change course
Maybe. The interest looks real, so the problem probably is too. Before changing the idea, we'd check whether it's the idea that's wrong or just how we explained it.
Test again
This is where we'd land. The interest is real and the understanding isn't. So we'd fix the explanation and run it again with new people. It's cheap, and it tells us which half was wrong.
Stop
Too early. 12 out of 40 is a real signal. Stopping now throws away the part that worked.
And the competitor? It tells us the problem is real. It doesn't tell us their answer is right.
Where AI fits in how we work.
The same rule as in our tools: AI does the heavy lifting, people make the call.
AI helps us
- Spot patterns across dozens of conversations
- Suggest angles and questions we missed
- Sketch mock-ups and prototypes in hours
- Turn raw notes into clear summaries
People decide
- What actually matters to artists
- What feels right, and what feels off
- How much weight a gut reaction deserves
- The decision itself
AI-made “users” sound convincing, but they're guesses. We never treat them as evidence. Real people only.
Traps we watch for.
Testing for show
Running tests, then doing what we'd planned anyway.
Hearing what we want
Picking the evidence that agrees with us. So we always ask: what would prove us wrong?
Asking about the future
“Would you use this?” gets polite answers. “When did this last happen?” gets the truth.
Falling for the first idea
We explore at least three genuinely different answers before choosing one.
Waiting for perfect
Endless research is just a slower way of guessing.
You're part of every step.
The Cyphermatix Community is how this works in practice: performers and producers who help us understand the problem, react to early sketches and test rough versions in their own setups.
- You'll always know who you're testing for.
- Partner projects are always your choice.
- Critical feedback counts as much as praise.
- You get a clear update on what changed because of you.
Working with a label, a promoter or a technology company? The route is the same: listen first, build small, test with real artists, decide on evidence. And the artists stay in charge of their part.
Not “What should we build?” But “What do we need to learn first?”
We always know our riskiest guess.
There's always a test running.
Every decision traces back to what people showed us.