Builders in review: The best takeaways from 2025 on our podcast

They show up in the quiet details. A hiring decision that went wrong before it went right, a team structure that only worked after someone admitted what wasn’t working, a moment of leadership that looked more like doubt than confidence.

In 2025, we sat down with leaders and builders in tech who are deep in the work: scaling organizations, navigating modern engineering challenges, adapting to AI’s rapid acceleration, and rethinking what “great talent” even means in a landscape that keeps shifting under our feet. Some conversations were tactical and immediately actionable. Others were the kind that stay with you for days because they change the way you see your own role.

This article is a look back at the best takeaways from this year’s episodes: the themes that kept resurfacing, the mindset shifts worth remembering, and the lessons that feel especially relevant as we step into 2026. Whether you’re leading a team, building a product, or simply trying to stay sharp in a noisy industry, we hope these insights give you something valuable to take with you.

Martina Johannesson, Unity: There’s always a risk of underwhelming users when introducing new tech

“We need the feedback, we need the usage, we want to ship it.” Martina Johannesson captured one of the biggest tensions in AI development: moving fast enough to learn, without moving so fast that users walk away after a bad first impression.

Because AI isn’t predictable in the same way as traditional product features, “perfect” is rarely a finish line you can wait for. But releasing too early comes with its own risk. Unity’s approach is a reminder that in creator-first industries, trust is fragile, and every AI launch has to balance experimentation with real expectations.

Annie Von Heijne, ICA Group: “Leaders need to stop wasting their teams’ potential”

Annie von Heijne put it bluntly: when you tell a data scientist exactly what to build, you’re not getting their best work. Instead, you’re reducing a problem-solver to an order-taker. The real value isn’t in executing instructions. It’s in understanding the problem deeply enough to challenge assumptions and find a smarter path forward.

And the cost of getting this wrong is higher than a suboptimal model. It’s wasted talent, slower learning, and teams that stop thinking proactively, something Annie reminds us we simply can’t afford.

Sepi Mohammadi, Apotea: The best decisions come from the ground up

Sepi Mohammadi shared a simple leadership lesson that many teams learn too late: as a tech lead, your job isn’t to blindly execute what comes from above. It’s to listen closely to the people doing the work. Engineers are often the first to spot what’s unrealistic, what’s missing, and what will break once it hits real users.

The strongest product decisions don’t come from the loudest voice in the room. They come from the ground up: from the people closest to the systems, the constraints, and the day-to-day reality of building something that actually works.

Sara Asgari, Telia: Cybersecurity has to be a leadership priority

Sara Asgari made the case for cybersecurity in the most practical way possible: every company will need to rely on data, analytics, and cloud services to move faster and stay competitive. That’s not optional anymore. But speed without protection is a risk most organizations aren’t ready to handle.

The real challenge isn’t just adopting hyperscalers or building analytics infrastructure. It’s knowing what to protect, what can live in the cloud, what must stay on-prem, and how to secure it all as complexity grows. In other words, if data is how you survive, cybersecurity is how you stay standing.

Ludvig Strand, Axel Johnson: AI sees the world through a Western, predominantly white lens

Ludvig Strand highlighted an uncomfortable truth: AI systems inherit bias by design, because they’re trained on human-made data. For LLMs, that often means a heavily Western, predominantly white internet. The result is that values and assumptions get baked into the output long before a company ever touches the tool.

Even when models are “aligned” or adjusted, it doesn’t erase the foundations they’re built on. As Ludvig bluntly put it, it can end up feeling like “putting lipstick on a pig.” The takeaway is clear: if you’re using AI across markets, you need to understand what it’s optimized for and who it might be leaving out.

Jacopo Himberg, Wolt/DoorDash: The Data Engineer title is losing its meaning

Jacopo Himberg shared how the data job market is shifting and how titles no longer reliably reflect the skills companies actually need. At Wolt and DoorDash, they stopped using the Data Engineer title entirely, because the role they were trying to hire for didn’t match what most candidates associated with it.

What they got instead were professionals highly skilled in SQL-heavy tooling (like dbt), while their biggest need was on the enablement side: building the data platform itself. In their model, the split is clear: some people work with data, others enable the work. And for that second track, they’re hiring mostly software engineers: a signal that hiring and job searching in data will increasingly be about capabilities, not labels.

Nima Salami, Oasys Now: Personalized medicine needs inclusive data and patient control

Nima Salami shared a mission that’s both ambitious and urgently needed: making personalized medicine accessible for everyone, globally. Today, much of modern medicine is built on clinical trial data that historically overrepresents white men, even though most of the world isn’t. That gap isn’t just unfair. It limits how well treatments work across different populations.

The way forward, Nima argues, is democratizing access to diverse, patient-driven data, and doing it with governance built in. By empowering patients to decide how their data is used through consent management, Oasys Now is working toward a system where precision medicine isn’t a luxury for a few but a scalable model that works for everyone.

Mina Boström Nakicenovic, Flightradar24: Growth isn’t hiring fast, but designing the organizational structure smartly

Mina Boström Nakicenovic reminded us that adding headcount isn’t the same as making progress. Rapid growth often leads to more layers, more complexity, and slower decision-making, which is exactly the opposite of what scaling is supposed to achieve.

At Flightradar24, the focus is on smart growth: shifting from relying on individual “local champions” to building consistently high-performing teams, while keeping the organization flat and horizontally scaled. Because when you avoid unnecessary hierarchy, you protect speed, clarity, and cost, and give great teams space to actually deliver.

Bryan DeNosky, Mojang Studios: AI is the worst it will ever be

Bryan DeNosky put the pace of change into perspective: AI is the worst it’s ever going to be right now. It’s improving fast, and the developers who learn how to work with it early will gain an advantage that compounds over time.

But the shift isn’t just using AI as “another tool.” It’s treating it like a collaborator: giving clear, verbose, pointed prompts and reviewing the output with real engineering judgment. In Bryan’s experience, it’s already reaching the point where AI can draft production-ready code, as long as the developer knows how to guide it and validate what ships.

Stefan Wendin, RISE Research Institutes of Sweden: We need to be honest about AI and job loss

Stefan Wendin doesn’t sugarcoat it: the idea that “AI won’t take your job, someone using AI will” is comforting, but increasingly incomplete. We’re already seeing entire functions being reduced or removed, not because individuals failed to adapt, but because systems are changing faster than organizations and people can realistically keep up.

Summary

This is only a snapshot of everything we covered on Builders in 2025, and we still had plenty of great conversations that didn’t make it into this year-in-review. If you want more real-world insights from the people building teams, products, and tech in the trenches, we recommend exploring the rest of the episodes.

And we’re not slowing down: new Builders episodes drop every Wednesday, so there’s always something new to learn, reflect on, and bring back to your own work.