The Founder's Story Tallinn by Google Cloud (October 8, 2025)
Visited local Google Cloud meetup - The Founder's Story Tallinn - organized by Jüri Kirpu & team at Kinomaja.
Priit Kaasik from Katana Cloud Inventory was a joy to watch. Casual and relaxed. He talked about personal history (php 👏 ), differences in company growth affected by US politics and its effects on supply chains, resulting in high cash burn rates and layoffs. But with adoption of AI and shifting the vision, they managed to get new rounds of investments. Wild startup ride. One takeaway is that you can speed up release cycles by just reducing pull-request review time by relying on AI. I think thats a fresh perspective, might be risky but desperate times call for desperate measures and it may actually be worth it. I certainly seen in various companies over the years cases where pull request gets so much opinionated feedback that it can take weeks to merge work. My question was about Katana potentially becoming a platform not just for managing inventory, and not just to understand objects better (a-la Mavenoid?), but for IoT / inventory talking to humans?
Marek Brazina from Google Cloud showed an evolution of LLMs going from pure text to images to now, agents. The talk led to AgentSpace, an environment to run agents.
I think A2A hype died out a little bit by now and I did not see that much autonomous agent adoption or collaboration. But maybe people are still catching up to workflows and agentic capabilities will reach us in the future. My question was about situation where agents could go into loops and the answer was that the platform needs to detect it and original agent is the orchestrator that needs to control task distribution. I think it sort of works with platform case, but if agents are implemented without Google, natively on local machine, I wonder how should we prevent agents spiraling out of control. Maybe a middleware with some counters, or maybe A2A defines that?
Panel
The panel was quite interesting.
Uku Tammet from Handhold touched a point of responsibilities that resonated with me. I think major difference between human employees and AI is the long-term role and responsibility associated with that. AI will just shrug, say sorry, you are right and make same mistake.
Taivo Pungas from Pactum AI wanted to see metrics (of incidents) that Priit Kaasik had. I liked the "Build something people want" mentality.
Martin Remmelgas from Codemagic was like a rebel and AI-skeptic among the crowd, which I liked. Disagreement was in where AI should NOT be used. Education. Mission critical things.
My question was should founders focus their time and attention on building anti-AI moat (if customers with AI can replace SaaS so easily) or should they focus of using AI as much as possible to be ahead of the curve and competition (with agents for example) and how find the balance between these. Answers varied from - focus on getting to 1M ARR to "Neither" and focus on product, to 70-20-10 rule (with AI being in the 10 area)
Hope we can see Google Cloud Estonia offices soon 💡