AI in the Mountains
Last week the Trice team got together for an offsite in the mountains of Norway. We spent 3 days in the snowy surroundings of Gausta and had time both for discovering the mountain, hanging out together and exploring AI and the latest agentic tools and technologies. We shared our experiences and experimented with new use cases and tools.
Our explorations and discussions covered mostly three tracks:
Rapid Prototyping With tools such as Lovable, you can rapidly build and app or website in just a few prompts. While these prototypes aren’t always production-ready, they are game-changers for exploring product and design ideas. The challenge we’re solving now is how to bridge the gap from a prompted prototype to a robust production system.
The Shift in Development Agentic development is inevitably the future. We shared some of our best tips and tricks on how we are currently working with it. We also discussed how this affects the role of developers, e.g. what happens with flow when the job includes much more context switching? The role is changing faster than any of us can really keep up with, what work will look like in just a couple of months is hard to say. It’ll take a few years before we really know the impact.
Agents in the Enterprise Understanding the capabilities and risks of agentic AI, specifically outside developer work. How will tools such as OpenClaw, Claude Cowork, Copilot Cowork etc fit into the enterprise? They come with a lot of security issues, but they are also becoming extremely powerful so solving these will open up a lot of possibilities.
One of our most interesting explorations revolved around securing AI agents by setting up a control plane around a sandbox-environment. When all communication goes via the control plane security measures can be put in place, making it completely safe to let the agent work in YOLO-mode inside the sandbox. We'll revisit this topic more in a future blog post.
