Insights

Architecture decisions, scaling patterns, and lessons from building AI systems in the real world.

LatestApril 15, 20265 min read

NIS2 Is Live: What Portuguese Companies Need to Know Right Now

Decreto-Lei 125/2025 took effect on April 3, 2026. Here's what changed, who's affected, and the compliance gaps most organisations don't see coming.

CybersecurityNIS2Compliance

2026

Apr 11

Apr 114 min read

Why Most Multi-Agent Systems Fail at Scale

Multi-agent systems aren't microservices. Here's what breaks at scale and the architecture patterns that survive production AI workloads.

ArchitectureMulti-Agent4 min read

Apr 9

Apr 94 min read

The Infrastructure Tax Nobody Talks About

Agent-generated deploys hit 30% on Vercel. The infrastructure layer is changing underneath us. Most companies are building on foundations not designed for this.

InfrastructureAI Agents4 min read

Apr 4

Apr 44 min read

Trustworthy Agents Are an Engineering Problem, Not an Alignment Problem

Anthropic's paper on trustworthy agents confirms what production teams know: agent reliability is an architecture problem, not a prompting problem.

AI AgentsReliability4 min read

Mar 25

Mar 254 min read

Vector Databases Are the Wrong Abstraction

Vercel open-sourced a knowledge agent using grep instead of embeddings. File systems beat vector databases for most AI agent workflows. Here's why.

ArchitectureRAG4 min read

Mar 18

Mar 184 min read

Green CI Is No Longer Proof of Safety

Agent-generated code passes every test and looks like senior engineer output. The old CI verification model is broken. Here's what replaces it.

AI AgentsInfrastructure4 min read

19 articles