Adapt Fast In The AI Era?

Adapt Fast In The AI Era?
Published on 10.07.2025

AI Agents are a hot thing these days. Everyone is integrating agents into their applications, processes and workflows. A clear trend is emerging: AI Agents are becoming more and more autonomous, and they are starting to take over tasks that were previously done by humans. This is a good thing, because it allows us to focus on the things that matter most, like spending time with our families and friends, or working on our hobbies.

In the near future, the machine to machine interactions will become more and more common. But currently the tooling is fragmented. You have n different agentic frameworks, each with its own API, and its own way of doing things. It's all quite chaotic, and disconnected.

Previously I wrote about speeding up the cycle time for delivering agentic software and I think the solution to this would be unification. Meaning allow easy deployment of tools and agents that are written in any language or any framework, provide agentic primitives like tooling, memory, sandboxing and so on, and allow agents to interact with each other in a standardized way with observability and monitoring built in.

This approach drastically reduces the time to market for agentic software, because you don't have to worry about the underlying framework or language. You can focus on the business logic of your application.

Companies can build complex agents for thier customers, individuals can build agents for themselves, and the whole ecosystem can benefit from this unification. This will make the feedback loop of building agentic software and measuring the impact of it's use much faster.

Faster cycle times, faster feedback loops, and faster adaptation to the changing world.

This is what the Agentic Internet is all about. It's not just about building software thats agentic, but also the underlying infrastructure that allows us to build and deploy agentic software quickly and easily.

I'm excited to work toward this vision, so I'm starting Concave.

If you wanna talk about this, DM me - @pwnfunction.