Rethinking the Cloud: Focus on Intent, Not Infrastructure

Right now, the cloud is a collection of resources abstracted into small pieces, accessible via APIs. But for most of us, it has become a nightmare of manual glue. We spend more time picking the right instance size, configuring VPCs, and wrestling with vendor lock-in than we do building our products.

The more I play with AI tools like Claude and Opencode, the more I realize we need to rethink the cloud entirely. We need to treat infrastructure the way we treat modern code: the logic and requirements are what matter, while the implementation should be invisible.

Here are a few principles for the cloud of the future.

Treat Cloud like Code

In software, we focus on the logic. In cloud, we are still stuck on the plumbing. We need a shift where intent is the primary driver. If you need a globally available, low-latency database, you shouldn't have to choose between RDS, DynamoDB, or Spanner. You should define the requirement, and a smart agent should provision, migrate, and optimize the resource automatically.

The Rise of the Agentic Infrastructure

The "Cloud of the Future" isn't a dashboard; it’s a living mesh of agents. To make this work, we need a few core components:

• Direct Resource Access: Agents shouldn't just suggest code; they should have the authority to deploy resources and make them public.

• Just-in-Time Databases: The right database should be selected and created only when needed. If your access patterns change from relational to document-based, the system should handle the migration for you.

• Default Autoscaling & Observability: We need a "Sovereign SRE"—a smart agent that tracks metrics and scales resources before the bottle-neck happens. If it can't fix a problem itself, it alerts you with a solution, not just a graph.

• Security as an Immune System: Instead of weekly scans, a continuous security agent should monitor syscalls and network traffic in real-time, isolating threats the moment they appear.

Is this feasible today?

For a long time, this was a pipe dream. But we are reaching a tipping point. To build this "Agnostic Control Plane," we need to bridge three gaps:

1. The Connectivity Gap: Using mesh networking (like Wireguard), we can treat AWS, local hosts, and edge providers as a single pool of compute.

2. The Physics Gap: We have to solve for latency. The AI must be "topology-aware," ensuring that compute and data stay physically close, regardless of the provider.

3. The Decision Gap: We need "Human-in-the-Loop" governance. The AI proposes the architecture and the budget, but the human clicks "Approve."

While the "Model Context" for these agents is still maturing, the underlying plumbing (Crossplane, eBPF, and Serverless) is already here.

Focus on Outcome, Not Output

The biggest benefit of an Agentic Cloud is that it aligns infrastructure with business value. Instead of asking "How many pods do we need?", you ask: "Is this € 100 server upgrade going to prevent a drop in conversions?"

Imagine a deployment flow like this:

• Developer: Pushes code and says: "Keep latency under 50ms for EU users and stay under €500/month."

• Agent: Evaluates providers, sets up a secure mesh, and provisions a database on a local provider because it’s 40% cheaper today.

• Sentinel: Scans the deployment for vulnerabilities and applies a patch before the site goes live.

Let’s build the future

The current state of cloud is too complex and too rigid. If we want to move fast, we have to stop being digital plumbers and start being architects of intent.

What do you think? Does your current cloud setup feel like a tool or a chore? If you’re interested in exploring how we can make your infrastructure more autonomous and business-driven, let’s connect. I’d love to brainstorm how we can turn your cloud into an outcome-driven engine.