Hey there,
2025 flew by. We grew from 35 to nearly 45 people and took on some technically complex projects, like a financial analytics system that took six months just to get the data models working right. We learned a lot about what we can actually build.
So here’s what happened last year and where we’re heading now.
Our biggest clients in 2025
This year we kept building with some great companies. Three stand out.
Xperi Inc. — a 3,000-person automotive and media company. We have 11 engineers working with them, and the partnership keeps growing.
Stevens Industries, Inc. — a 70-year-old manufacturing company with over 1,000 employees. 13 of our team members are helping them build ML and AI-powered estimation tools that turned hours of manual work into minutes.
Internet Archive — a digital library archiving websites, books, and media. We’ve got 3 engineers supporting them.
We’re grateful for these partnerships, and we’re planning to grow alongside them.
AI in woodworking: what we learned
We surveyed 40 woodworking professionals about AI adoption in their industry. The results?
Half use digital tools: software for design, estimating, and project management. But only 1 in 5 have actually started using AI. There’s a big gap between “we’re digitised” and “we’re using AI to automate work.”
The top challenge remains the same: estimating and takeoffs — figuring out how much material a project needs and what it’ll cost. Over a third of respondents said it’s their biggest time drain. And when we asked what they want to learn about AI, 60% said: can AI actually read architectural drawings and do accurate takeoffs?
The answer is yes. We’ve built systems — like the one for Stevens Industries — that take estimation time from 2 to 6 hours down to under 10 minutes.
One shop from our survey is already doing this: they’re using AI to read architectural prints and pull out specific information, saving 30 to 60 minutes per estimate. It works. But they still won’t use it on their biggest projects. Why? They’re not confident it’s accurate enough to trust with a $500K bid.
If you want to dig deeper, you can read the full survey results here.
Heading to Switzerland in 2026
Four projects came from Switzerland in 2025. Two things stood out.
First, all of them wanted AI integrated into existing systems, not new products built from scratch. Companies with established CRMs and ERP platforms realise their products feel outdated without AI features, but their teams lack the expertise to build those integrations. We added AI engineers to their teams to bridge that gap.
Second, one project involved training a model specifically for Swiss German. General LLMs handle Swiss German poorly: it’s not a dialect they were heavily trained on. So we’re training a custom model on Swiss German datasets for language recognition. General purpose LLMs won’t get to these local dialects quickly, so specialised models still matter.
So I’m booking a business trip to Switzerland. I’ll research the market, meet existing and potential partners, and attend CloudFest, one of the largest cloud events in Europe, held not far from Switzerland in Germany’s Europa-Park. Also, I’ve been told that the chocolate in Switzerland is pretty good.
If you’re going to CloudFest — I’ll be there for the third time — let’s talk.
AI Audit: a new service in 2026
Vibecoding is having a moment, and for good reason. You can go from idea to working prototype incredibly fast. Perfect for testing concepts and getting buy-in.
But prototypes and production systems are built differently. A demo handles 10 users perfectly. Production needs to handle 10,000, and that’s when you discover your database can’t scale, your prompts are inconsistent, and nobody documented how anything works.
That’s when we usually hear from them: “How do we make this production ready?” “Should we rebuild or can we scale what we have?”
So we’re launching a new service: AI Audit. Our AI architect and head of delivery review what’s been built, identify what’s salvageable, and create a roadmap for turning a prototype into a real system.
Top request in 2026: AI integration
Hard to miss how much demand there is right now for integrating AI into existing systems. Not building from scratch, just adding AI features where users already expect them.
Your customers are using ChatGPT. They’re using Claude. They expect software to have AI capabilities now. If your product doesn’t, it feels outdated. We’re helping companies bridge that gap, bringing AI into legacy systems without ripping everything out and starting over.
Workation in the Canaries
Most of my time I work from my home office in Lisbon. There’s so much work that I rarely make it out of the house during the week.
One way I deal with burnout and that stuck feeling: workations. A few times a year, when I realise I’m hitting a wall, I book a ticket and work from somewhere else.
In February, it was the Canaries. I spent the week working and thinking through our direction for 2026 — planned conference visits, started mapping out a US trip for the autumn. I was there four years ago with friends but never drove around the whole island. This time I did, took the whole day, stopped at different viewpoints.
Working from a different place doesn’t solve everything. But three decisions I’d been putting off for months became obvious that week.
What’s next
We’re growing the team and shifting focus: less outstaff, more outsource. Less “here’s an engineer for your team,” more “here’s the complete solution you actually need.”
Discovery phases, AI/ML solutions, and helping companies move from “we think we need AI” to “here’s exactly what we’re building and why” — that’s where we’re doubling down.
That’s 2025: three incredible clients, 40 woodworking shops surveyed, four Swiss projects, one new service launching, and a lot of conversations about whether vibecoded prototypes can actually scale. Spoiler: they can’t.
That’s the year. Now let’s see what 2026 actually brings.
See you around,
Volodymyr
This issue first went out to subscribers of our LinkedIn newsletter, AI Era Development Stories, on March 11, 2026.
