Software Mansion

Swap unit tests for higher layer tests. I’ve been relying more and more on e2e over unit tests for years and I feel validated now. Agents no longer make human-like mistakes (like using wrong variable name randomly) – any mistakes they make are more in tune of making the wrong choices when the prompt is too vague, or bloating abstraction layers. I feel e2e gives me almost all of value without increasing the codebase and duplicating logic in code and unit tests. Good comprehensive e2e tests help avoid regressions, and it's really easy to cause regressions now, with the increased amount of code shipped and less and less time spent on review.

Partiful's Android app now matches iOS. Here's how we worked with their team to get there.

Need a few hours away from your desk? Bring your laptop, your half-finished repo, and your questions to our event. We'll show you what on-device AI can achieve, you'll get to check out some newest local models and have some good food with other Krakow builders.

If you're building something with AI in Krakow, this is an event for you. You’ll get a space to tinker on your laptop, talk shop with people who actually get your problems, and see how on‑device AI can slip into your real projects.

LLMs are getting more and more expensive, while on-device AI is getting better and better. Check out the newest models, play around with our demos and spend the day with other builders from Krakow.

Come work with other builders from Krakow, play around with new local AI models and get inspired at our AI Brunch! We're opening our office for a day, so you can code, network and check out what on-device AI can actually achieve when you give it a chance.