Building Through Uncertainty: A Conversation on Resilience, AI, and the Future of Software (feat. Asia Solnyshkina)
A founder and product strategist discuss building software through uncertainty, AI's impact, managed services, vibe coding, hiring, and global perspectives.
A conversation between Krish Palaniappan and Asia Solnyshkina, founder of ProSense Digital — exploring what it means to build software in an era when the rules are being rewritten in real time.
Podcast
Who Needs Developers? (Everyone, Actually) - on Apple and Spotify.
From Moscow to the World: A Founder’s Journey
Asia Solnyshkina did not plan to become a global founder. In March 2022, when war broke out between Russia and Ukraine, she left Moscow with little more than her laptop and her two children. There was no plan, no destination, no certainty about what would come next.
“I left Russia with just my laptop and two kids,” she recalls. “Basically with no plan of what I will be doing, how I will be managing my company.”
What followed was a journey across continents. First Georgia, then three years in Mexico City — “a brilliant, beautiful place” — and now an attempt to settle in the United States. Her company, ProSense Digital, builds custom software for businesses worldwide: ERP systems, CRMs, websites, and complex automation tools. Clients range from the United States to Latin America to Australia.
The eight-year-old company was remote from the start, which softened some of the disruption. But rebuilding a business across countries, time zones, and cultures forged something more durable than any office could: resilience paired with agility.
“I do feel that right now I can adapt to the new world a lot,” she says, “because I’ve been traveling, I’ve been meeting different people, I’ve been working with different businesses, rebuilding the whole structure, the whole company, losing partners and all of this. I feel comfortable in this new AI era, which is pretty fast.”
What Good Design Actually Means
Working across Russia, Latin America, Australia, and the US revealed striking differences in how clients approach software. In Russia, Asia found, clients often arrived focused on aesthetics — the pretty button, the beautiful interface — sometimes at the expense of the underlying system.
“What I felt about building business in Russia was, it was all about, let’s do the pretty UI. And that’s it. We’re not thinking about the UX,” she explains. Her work became as much about education as engineering: helping clients see that beneath every button there must be a system that serves a real business goal.
This is where the conversation got pointed. What does “good design” actually mean to engineers who want specifics, not adjectives?
For Asia, the answer is unromantic: good design is design that converts. Amazon, with its dense interface and relentless commercial focus, is good design. Award-winning agency sites with floating parallax and ornate animations often are not. “I’m not thinking about how beautiful it is, I’m thinking about the goals. I’m thinking about what people are trying to achieve.”
That definition is debatable — and Krish pushed back. Plenty of well-designed products fail to find product-market fit, derailed by timing, capital, or distribution rather than craft. But the underlying point holds: design exists to serve business outcomes, not to win Dribbble shots.
The AI Inflection: Cheaper Software, More Software
A reasonable prediction, repeated for years, holds that AI will end software development as a profession. Asia’s lived experience contradicts it.
“I’ve been told for years in a row that software development will be dead like in a year, in a month or so. Right now what I’m experiencing with my exact business — it’s actually not just thriving, but my client base grew.”
This echoes the Jevons paradox: when something becomes cheaper to produce, demand often expands rather than contracts. Software is following the pattern. Businesses that once viewed custom development as expensive and slow now see automation as accessible — and they are bringing more problems to the table than ever.
Asia’s design and prototyping process has changed dramatically. Where her team once spent weeks in Figma, iterating through three or four rounds before showing clients anything tangible, they now prototype directly in tools like Lovable. By the time the polished design would have arrived under the old process, the market itself might have shifted.
“The main essence of what we’re doing is prompt engineering,” she says. “Creating a good task for AI so it could understand the problem we’re trying to solve. Not drawing beautiful buttons, but solving the real business problem.”
The Managed Services Question
If anyone with a credit card and a Lovable subscription can build software, what is a managed services provider actually selling?
Krish pressed on this directly. The traditional moat — knowing a particular language, framework, or architecture — has weakened as tools generate working code from natural language. So what does Asia’s company offer that a curious non-engineer cannot do alone?
Her answer pivoted away from the tool entirely. “I’m not using just the tool, because the tool is just the tool. I’m using my experience working 15 years in software development.” More importantly, she argues, ProSense Digital is not a body shop selling hours — it is a product company selling outcomes.
“We’re not trying to sell just the lines of code. We’re trying to sell the complete products that helps people with whatever they need.”
What AI changes for her company is leverage, not category. Experiments that once cost real money — A/B tests, prototype iterations, market probes — now cost almost nothing. That makes it easier, not harder, to do the thing she has always sold: understanding what users actually need versus what they say they need.
The honest concession: she may be wrong. “Probably in a year or two, I will have to go to some other business. But right now I do feel like this. We’re building products.”
What Founders Get Wrong
Asked what founders most often get wrong when scaling, Asia gave an answer rooted in the cost of conviction. Founders fall in love with their original idea and refuse to let market signal change their minds. “Sometimes founders stick to their ideas even though they are in the process of developing the product itself, they do understand that probably this idea is not right. But they’re investing a lot of time, a lot of money and a lot of everything.”
The discipline she advocates is experimentation as default. Talk to users, watch behavior, run tests, and accept that what people say they need is rarely what they actually need.
The conversation circled into a productive disagreement here. Krish raised the Henry Ford line — that customers asked for faster horses, not cars — and the iPhone launch, which Asia herself remembers as underwhelming at the time. Sometimes great products are not validated by initial reception. Sometimes the surveys say no and the founder presses on anyway.
The synthesis: even the giants get this wrong. Meta’s Metaverse spend, Google’s graveyard of canceled products, and the cool reception to Meta’s smart glasses all suggest that even well-resourced teams build products for ego, for investors, for narrative — not always for users. Asia’s framing: “Sometimes people are building products not to be successful.” It is a sharp observation about R&D, ego, and the pressure to appear ahead.
The Future of Software Development
Krish offered a candid read on his own field after more than two decades in it. Software has never been static, but the pace of change in the last two years is different in kind, not just degree. Several things have shifted:
The fundamental shift is that he no longer needs a developer to build software. After 20 years of always needing one, that assumption is gone.
Hiring is harder to think about, not easier. Yes, anyone can use these tools. But if a hire cannot reason from first principles about persistence layers, about why Postgres versus DynamoDB, about architecture trade-offs — then what value do they add beyond what the model already provides?
Architecture itself is changing. Engineers with muscle memory from the previous era have to actively unlearn old patterns. Newcomers have an advantage in flexibility but lack the scar tissue that distinguishes good decisions from bad ones.
The economics are commoditizing. Charging top dollar for code is over. Smaller teams shipping more software is the emerging shape. Founders report going from 54 people to 8. Yet layoffs are everywhere, and the gap between “AI made us more productive” and “we still have headcount” is closing painfully.
Production reality is more complicated than the demos suggest. AWS suffered outages that the company attributed to AI-generated code; senior architects must now approve generated changes there. Apple is rejecting vibe-coded apps from the App Store. Vibe coding is excellent for experiments and prototypes — Asia uses it actively — but production-grade systems still demand engineering judgment.
“I’m not comfortable pushing code to production that I’ve at least not seen one time,” Krish said. “I cannot have a tool generate code and then push it to production.”
Hiring in the New World
Asia’s hiring philosophy has quietly evolved into something unconventional. She does not run formal interviews. Instead, every manager keeps a stockpile of small, low-stakes tasks — the kind where a candidate failing would not damage anything important. When a CV catches her eye for curiosity and intelligence, the candidate gets one of those tasks.
“I’m observing how they are interacting in the real world setting.”
College degrees are not required. The trait she screens for, above all else, is curiosity — the willingness to engage with a world that is changing faster than any curriculum can keep up with. She is actively hiring vibe coders, not because they replace engineers, but because they extend her ability to run cheap experiments at scale.
On Jobs, Identity, and What Comes Next
Krish was direct when Asia asked whether he feared AI would take his job: “I’m not afraid because I know it is going to. I have no doubts about that. The job that I have done all these years — writing code, like every line of code — that job is gone. It’s not coming back.”
The dilemma is more subtle than replacement, though. Sometimes he sits down to write a line of code and hesitates because the tool can do it. Then the tool’s output is not quite right, so he rewrites it. At which point, why didn’t he just write it himself? The judgment about what to delegate and what to keep is a new skill, and it requires the engineering background he’s not yet willing to abandon.
“You want to use these tools to make yourself more productive, but I don’t want to use those tools to lose my agency. We are all born with a certain intellect, good, bad, or ugly. If you don’t end up using that, what is the point in living life?”
Asia’s view on AI’s broader employment impact is more optimistic. Yes, jobs will disappear. But new ones — for people who can think, adapt, and stay curious — will emerge. The transition rewards people who treat this as a moment to experiment, not a threat to defend against.
The Future of Managed Services
The managed services model for custom software development is undergoing a fundamental structural shift driven by AI-assisted code generation and rapid prototyping tools like Lovable. Historically, the value proposition of firms like ProSense Digital rested on deep technical expertise in specific stacks — React.js, Python, PHP — and the human capital required to translate business requirements into functional ERP or CRM systems over multi-month development cycles. Today, that cycle has compressed dramatically. Rather than spending two to three weeks on Figma prototyping followed by iterative design reviews, teams can now generate working UI prototypes through prompt engineering in a fraction of the time. The core competency has shifted upstream — away from implementation fluency and toward problem framing, requirements elicitation, and knowing what questions to ask the machine. Companies that continue to sell lines of code as a deliverable will face severe margin compression; those repositioning around product outcomes and experiment-driven iteration are finding, counterintuitively, that demand is actually growing as the Jevons paradox plays out: lower build costs are expanding the total addressable market for software.
The architectural risk introduced by vibe coding and LLM-generated codebases is becoming increasingly visible at scale. AWS’s recent production outages, attributed in part to AI-generated code reaching production without sufficient senior review, illustrate a critical gap: the speed at which code can be synthesized now far outpaces the institutional knowledge required to validate it. Key engineering decisions — selecting appropriate persistence layers (e.g., PostgreSQL vs. DynamoDB), designing for idempotency, managing stateful distributed workflows — require understanding that is not easily delegated to a generative model. Apple’s App Store rejections of vibe-coded submissions further underscore that AI-generated code often fails production-readiness criteria around security, performance, and platform compliance. The practical implication for engineering organizations is a bifurcated workflow: use AI-assisted generation aggressively in prototyping and experimentation phases, but ensure a senior architect with domain fluency reviews and approves anything moving toward production. The engineering background requirement hasn’t disappeared — it has simply migrated from writing code to governing the code that machines write.
A Final Thought on People
After a wide-ranging conversation, Asia closed with the observation that surprised her most across years of travel and work in Russia, Latin America, China, Singapore, Australia, and the US:
“All of us are people. We all are little children inside. We all have the same fears, the same joy, the same everything.”
Business cultures differ. Design preferences differ. Management styles differ. But under all of it, the people are remarkably the same — and the work, in the end, is for them.
Asia Solnyshkina is the founder and CEO of ProSense Digital, a product company building custom software for clients worldwide. Connect with her at prosense.digital or on LinkedIn.