Sometime last month I opened the VALK Figma file looking for a component I built in 2020. I needed to check how I’d handled a particular empty state, one of those obscure screens that appears when an investor has no active deals. I found it eventually, buried somewhere around frame 1,400. The page has thousands of frames now. I remember when it had forty.
That empty state? I’d designed it in three hours one afternoon in Alanya, Turkey, sitting at a rental apartment desk with a ceiling fan that made too much noise. A throwaway decision. Grey illustration, short line of copy, a button. Three years later it’s still there, unchanged, shown to asset managers at seventy-something financial institutions. I stared at it for a few minutes and felt something I couldn’t name exactly, not embarrassment, not pride. Something more complicated than either.
Four years. Same product. Same Figma file. Same codebase underneath everything I’ve drawn.
What agency work taught me (and what it didn’t)
Before VALK I spent two years at Spacecode, a Moscow agency. In that time I touched something like fifty-plus projects, banking apps, security dashboards, enterprise tools, consumer products. We’d sprint into a new domain every few weeks, learn enough to design something coherent, hand it off, move on. Every Monday felt like a new puzzle.
I got very fast at Spacecode. That’s the main thing. Fast at reading a brief, fast at identifying the shape of a problem, fast at generating concepts. Speed was survival there. A week of slow output had real consequences; other projects were waiting, other clients were calling.
What the agency years couldn’t teach me was depth. You can’t develop depth in six weeks. You understand the surface of a problem, sometimes the layer below it, but you rarely get to the ground. You hand off your designs and you never find out if the empty state you designed in three hours becomes a permanent fixture that thousands of people see every year.
At VALK, I found out.
The weight of accumulated context
The thing about staying with one product for four years is that you know it the way you know your own apartment. Where the awkward corner is. Which drawer sticks. Every workaround that got quietly institutionalized.
I know every edge case in the VALK platform, not from documentation, but from having created the edge cases, discovered them, patched them, or consciously decided to live with them. I know which design decisions were made under deadline pressure in 2020 that became load-bearing by 2021 when we scaled to dozens of institutions. I know which navigation pattern was “temporary” and is now four years old and woven into how every user expects to move through the product.
This kind of knowledge is genuinely useful. When someone proposes a change to the deal management flow, I can trace second and third-order effects almost automatically. I can say “if we move that action here, it creates a conflict with how compliance users approach the same screen from a different entry point”, and I know this not because I ran a workshop, but because I lived through the version where it was different and watched what happened.
The weight of that context, though, is real. Every new design decision carries the shadow of every past one. Early on I could move quickly because I didn’t know what I didn’t know. Now I know too much. I know which assumptions are fragile. I know which constraints are genuinely immovable and which ones just feel that way because they’ve existed for two years. Making confident decisions has become harder, not easier.
I don’t think most design writing is honest about this. There’s an assumption that experience makes you faster and more decisive. It can. But with enough context, it can also make you slower, because you genuinely understand the complexity you used to move through without noticing it.
The “quick fix” that never left
In early 2020, we had an urgent issue with how deal documents were displayed in the investor portal. The hierarchy was broken, the most important documents were getting lost. We had a release coming. I designed a fix in an afternoon, a card pattern with status indicators I’d never used in this product before, borrowed from something I’d built at Spacecode.
It worked. The release went out. Nobody complained.
By the end of 2020, that pattern had propagated into four other sections of the platform. By 2021, when financial institutions were onboarding rapidly, it was part of the visual language new users learned first. By 2022, it was referenced in our onboarding materials.
I never formally chose that pattern. I chose it under pressure on one specific afternoon, for one specific problem, in a city I’d only been in for a few months. It’s now one of the most recognizable UI elements in a product used by banks and hedge funds across fifteen countries.
There’s no such thing as a temporary decision in a living product. Whatever you ship becomes the baseline. The longer it survives, the more weight it carries, until removing it feels like surgery rather than editing.
The loneliness of being the only designer
Something I rarely write about directly: in four years at VALK, I’ve been the only designer. No one to compare notes with, to catch my blind spots, to share the particular frustration of caring about UX in an organization full of people who care about other things.
At Spacecode there were other designers around. Not always on the same project, but present. You could ask someone to look at a flow. That kind of creative friction is valuable in ways that are hard to appreciate until it’s absent.
At VALK, the product vision lives almost entirely in my head. Mostly that’s fine, I’ve made peace with it, found ways to create useful friction through documentation, through forcing myself to write out rationale rather than just designing it. But there are moments, usually when I’m sitting with a difficult problem that has no clean answer, where I’d give a lot to have another designer in the room. Not to validate me. Just to think alongside.
The Figma file with its thousands of frames is a conversation I’ve been having with myself for four years. That’s its own kind of loneliness.
Patience and restraint, the skills I didn’t know I needed
Agency work rewards output. The artifact matters because the client is evaluating the artifact. The velocity is visible and measurable.
Product work rewards restraint. The best decisions I’ve made at VALK in the last two years are things I didn’t do, flows I didn’t redesign, patterns I didn’t change, features I pushed back on. These don’t show up in a portfolio. There’s no frame in Figma that says “Dmitry resisted the urge to redesign this for the third time and the product was better for it.”
When you’re redesigning something that real users with real professional responsibilities depend on, the cost of disruption is higher than the cost of imperfection. You learn to sit with imperfect solutions while you gather enough evidence to justify changing them. You learn the difference between something that’s wrong and something that just bothers you.
That second category is bigger than I expected.
Restraint is harder than it sounds. Especially when you’ve been looking at the same interface for four years and you can see every flaw. You know you could make it better, but “better” has to be weighed against the transition cost to people who’ve built workflows around the current version. That calculation doesn’t exist in agency work. In agency work, better is always worth it, because nobody has a workflow yet.
When variety is a feature, not a distraction
I’d be lying if I said I don’t miss the agency rotation sometimes. The context-switching was exhausting but the variety was genuinely stimulating, every new project a new vocabulary to learn.
Here in Los Angeles, I sometimes catch myself scrolling through design work online and feeling something like nostalgia for problems I’ve never actually worked on. Healthcare UX. Education platforms. Tools for creative professionals. I’ve spent four years getting deeply fluent in one domain, private capital markets, asset tokenization, institutional finance. That fluency is real. But it’s also a kind of narrowing.
Depth and variety pull in opposite directions. You can get both if you move between deep-product roles over time, but within a single job, the longer you stay, the more specialized you become. I know this product. I do not know what I would do with an unfamiliar problem the way I did in 2018. Some of that flexibility has been traded for expertise.
Whether that’s worth it is a question I’m still working out.
What four years actually builds
There’s a kind of satisfaction in long-term product ownership that has no equivalent in agency work. When I watch a financial institution use a flow I designed two years ago, watch them navigate it efficiently, without friction, as if it’s obvious, that’s different from watching a client approve a prototype. The approval is gratifying. The actual sustained use is something else. Evidence that you built something real.
I’ve designed for people who have real professional stakes in the product working correctly. Not user-testing participants. People who depend on this software to manage actual capital. That responsibility has made me a more careful designer than any brief or deadline ever did.
Agency taught me speed. Product taught me discipline. I needed both, but I had to learn them separately, and I couldn’t have told you that in advance.
A question I keep coming back to
In the last few months I’ve started to notice a thought that keeps returning. Not anxiously, but persistently.
How many more years of this?
Not because I want to leave, I’m genuinely invested in what VALK is building. But four years is a long time with one product, and I find myself wondering when deep expertise tips into insularity. When does caring deeply about one product become a limitation on your growth as a designer?
Some designers spend a decade with a single product and produce extraordinary work. Others need the rotation to stay sharp. I’m still figuring out which kind I am, which feels like an odd thing to still be working out at four years in.
The Figma file keeps growing. The empty state from 2020 is still there, ceiling fan and all.
Originally published on kargaev.me. Imported to blog.deeflect.com archive.