Writing from Luiz Ozorio on AI systems, agent architecture, LLM infrastructure, observability, evals, and supervised control systems for reliable intelligent software. https://cloudpresser.com/writing 2026-05-11T00:00:00.000Z https://cloudpresser.com/writing/why-coherence-doesnt-scale-with-capability 2026-05-11T00:00:00.000Z 2026-05-11T00:00:00.000Z

There's a quiet assumption underneath most AI discussions: if capability keeps improving, coherence will eventually follow. Gödel's incompleteness theorems suggest otherwise. The supervision layer isn't a temporary workaround — it's part of the design.

https://cloudpresser.com/writing/why-ai-needs-control-surfaces 2026-04-27T00:00:00.000Z 2026-04-27T00:00:00.000Z

Once you can see what the system is doing, the next problem is interacting with it. Chat interfaces for AI agents are like flying a drone through a text terminal. The industry needs purpose-built control surfaces.

https://cloudpresser.com/writing/observability-for-ai-agents 2026-04-20T00:00:00.000Z 2026-04-20T00:00:00.000Z

AI systems fail in ways that look like success. You can't find these failures in a chat log. You need traces. AI agent systems need the same observability infrastructure that distributed systems built over the past decade.

https://cloudpresser.com/writing/ai-agents-are-control-systems 2026-04-13T00:00:00.000Z 2026-04-13T00:00:00.000Z

We're building AI systems like they're chatbots. They're not. They're control systems. The architecture that robotics solved decades ago — machine, telemetry, interface, human — is the same architecture AI agents need.

https://cloudpresser.com/writing/smart-model-reviewer-is-backwards 2026-04-06T00:00:00.000Z 2026-04-06T00:00:00.000Z

Most AI pipelines have generation and verification backwards. Smart models should generate. Cheap models should verify. The industry is putting its best capability in the wrong place — and it's creating a quality ceiling, not just a cost problem.

https://cloudpresser.com/writing/bash-is-all-you-need 2026-03-17T00:00:00.000Z 2026-03-17T00:00:00.000Z

The industry phrase 'bash is all you need' is technically correct. But it hides the real problems: context management, execution reliability, alignment with outcomes, and human supervision.