AI and software are rapidly changing and show no obstacles in its continued growth. For builders, businesses or founders it's become clear that the highest leverage task you can do is think about the future.
At Subterranean we've made a couple future bets already, all stemming from the idea that we wanted to support whichever shape would make AI agents the most useful but also reliable and secure. In just the past few months, we've seen agents excel at all kinds of roles and become indispensable to the companies that will dominate future markets.
First, it helped lower the barrier for people interested in apps and software to start building them. Coding was always a technical obstacle in order to build, but now that's largely commoditized by frontier models that live and breathe in programming languages and markdown.
If coding tasks can now be accomplished in one try, what's stopping prospective builders? The age-old challenge of designing and building a great product. Whether it's a mobile app or enterprise SaaS platform, all products need to be deliberately and carefully designed by someone with taste and wisdom. Without this guidance, junior engineers or AI alike will continuously tweak UI and add features without ever building something that actually attracts and retains users.
This is when we realized that using agents is actually a two-way relationship. You don't just use an agent to get something done. Rather, the best users of agents are actually collaborating directly and constantly tweaking and updating what makes up their agent. New workflows are cemented into skills, new tools are connected through MCP, and emerging roles and responsibilities are built into new or existing agent SOUL.md files as agents become a bigger part of one's org.
Where is this all going? Many market and tech experts like to describe the tech and AI world "converging" into agent-driven architecture. It means a new world where agents are the primary workers who interact with tools and communicate on the internet. They're the "hands" while humans are the "brains" behind the scenes, managing and delegating.
There's still many many pieces missing before this future arrives. In order to have a nearly full trust contract with our agents, we need reliability, security, governance, and all of the tools to be able to observe and evaluate how they actually perform. We need to be able to deploy agents into the real world, running independently 24/7 on cloud infrastructure with all of the internet and computer tools it needs. We need to release AI agents out of our local hardware and into workspaces that were actually designed for them, making it easy to write and retrieve context of all kinds.
We're tackling each of these pieces in our new platform.
I'm excited to share this journey and experiment together.
Gordon