Building taste through experiments

Taste is not something you have

There is a version of the word "taste" that functions as a credential. You either possess it or you don't. It belongs to people who were exposed to the right things early: the right schools, the right cities, the right rooms. By this account, taste is essentially inherited. It's a feature of upbringing, not effort.

That account is wrong, and the wrongness of it has real consequences. If you believe taste is innate, you stop trying to develop it. You defer to people who seem to have it. You treat it as something mysterious, present in certain people, absent in others, and you exempt yourself from the slow, uncomfortable work of building it.

The alternative view is less flattering but more useful: taste is built through a long sequence of experiments, each of which teaches you something, and none of which teaches you enough on its own.

This has been the working assumption at Mustard Seed Group since the beginning, and May 2017 feels like a reasonable moment to make it explicit.

What an experiment actually is

The word gets misused. In most product contexts, "experiment" means an A/B test. You run variant A against variant B, measure a conversion metric, and ship the winner. That is useful, but it is not what we mean.

An experiment, in the sense we care about, is any act of making something and then honestly evaluating what came out. The evaluation is the hard part. Most people make things and then defend them. The rare skill is to make something, sit with it, and see it clearly: what it does well, what it fails at, what assumption it reveals, what direction it points toward.

You cannot shortcut the making. The evaluation is only meaningful if you have genuinely committed to the thing: if you built it to your honest best understanding, not to a safe minimum. A half-committed experiment teaches you almost nothing, because you can always blame the output on the effort rather than the idea.

And you cannot shortcut the honesty. This is where most people get stuck. There is a deep human pull toward protecting what you have already made. The portfolio is only as good as the group's willingness to evaluate its own work without flattering itself.

How you build taste through this

The mechanism is not mysterious. You make something. You look at it. You notice what is off: the proportion that doesn't quite work, the copy that is trying too hard, the interaction that requires more explanation than it should. You try to articulate why it is off. You make the next version with that articulation in mind. The next version surfaces a different set of problems. You articulate those. You keep going.

Over enough iterations, two things happen. First, you develop a vocabulary for what you're noticing. Problems that were once vague, "something feels wrong here", become specific: the hierarchy is unclear, the pacing is uneven, the call to action is earning distrust rather than confidence. Naming the problem is most of the solution.

Second, your instinct sharpens. Early on, you need to run through the checklist. Later, you see the problem immediately. That instinct is not magic: it is pattern recognition built from repeated, honest evaluation. The evaluation had to happen for the pattern to form.

This is the feedback loop that builds taste. It requires volume: you need enough experiments that the patterns can emerge. It requires honesty: you need to see the work as it is, not as you wished it was. And it requires time: there is no compressed version of this process that produces the same result.

The three kinds of taste that matter here

Taste, broadly defined, covers a wide surface. But within the MSG portfolio, three distinct kinds show up most frequently.

**Product taste** is the ability to judge whether something works as a product: not whether it is technically sound, not whether users can figure it out with enough effort, but whether it earns its place in someone's workflow or life. Product taste is what you exercise when you cut a feature because it adds cognitive load without adding value. It is what you use when you decide that a simpler version of the thing, less polished, fewer options, actually does more for the person using it. Orbit is where this gets tested most directly. A B2B operating surface that covers the full commercial workflow is a sprawling thing to build. The only check on sprawl is a consistent sense of what the product should and shouldn't do, which requires product taste developed through building and evaluating many versions.

**Creative taste** is the ability to judge whether something communicates: whether the thing you have made lands with the person it was made for, in the way you intended. This operates in visual design, in writing, in the structure of a conversation, in the rhythm of an interface. Naira, as a performance coach inside CheekyGains, is partly a creative challenge. An AI coach that communicates poorly is worse than useless. It actively undermines trust. Getting the voice right, the timing right, the level of directness right: these are creative decisions, and developing the taste to make them well requires a body of creative experiments behind you.

**Commercial taste** is the ability to judge which ideas are worth pursuing at all: not just whether they work in principle, but whether they will hold up in a real market with real constraints and real alternatives. This is the hardest to develop because the feedback loop is slowest. You make something, you take it to market, and months pass before you know what you're actually dealing with. TUXX operates in this space most directly. Building custom AI systems for clients means making judgment calls about what patterns are worth productising, what requests are genuinely commercial versus one-off, what the difference is between a client who will stretch you and a client who will dilute you. These calls require commercial taste, and commercial taste requires having made a lot of bets and lived with the outcomes.

Why the portfolio benefits from all three

These kinds of taste are related but not identical. Someone with sharp product taste may be a poor judge of commercial viability. Someone with excellent commercial instinct may not be able to see why a design decision undermines the product's trustworthiness. And creative taste, without the others, can produce work that is admired but not used.

The case for running multiple products across multiple categories is not just about revenue diversification. It is about cross-pollination of feedback. What the creative work in All Purpose teaches about communication and rhythm feeds into how Orbit surfaces information. What the commercial reality of TUXX client work teaches about what organisations actually need informs what Benediction Lab chooses to research. What Benediction Lab discovers about agent behaviour and memory systems becomes material for what Orbit can eventually do.

The experiments are not isolated. They are running in parallel across a portfolio, and the feedback from each one is potentially relevant to all the others. The compound effect of this, if the evaluation is honest and the connections are made, is a group that develops taste faster than any single-product studio could.

The discipline this requires

None of this runs on goodwill alone. The experiments only teach what they teach if the evaluation happens with some rigour. This means having clear criteria for what you were trying to do before you assess whether you did it. It means separating "I don't like this" from "this doesn't serve its purpose." It means sitting with failures long enough to understand them rather than discarding them as quickly as possible.

It also means accepting that most experiments will not work as hoped. This is not a failure of the process: it is the process working correctly. The experiments that go wrong are often the most instructive. The version that nearly worked but missed in a particular way tells you something specific about where your judgment was off. The version that failed entirely tells you something about your initial assumptions. Both kinds of information are valuable. Neither is available if you don't make the thing.

The working thesis at MSG has always been that capability is built, not found. That is as true for taste as it is for anything else. The difference between where we are and where we want to be is not a matter of waiting for the right moment or the right collaborators or the right market conditions. It is a matter of making things, seeing them clearly, and adjusting. Repeatedly, and with patience.

That's the work.