Parallel isn't coordinated: the real bottleneck in agentic coding
“Analysis is running in parallel (4 agents)…”
agentic-ai
A cloud, platform, and security architect writing about how large systems actually get built, secured, and adopted.
Platform Engineering
DevSecOps
IAM & Identity
Software Supply Chain
Cloud-Native Architecture
Observability
Agentic AI
Recent posts
Your AI agent doesn't have an ID. It borrows yours.
Here’s a small, uncomfortable fact about the AI agents everyone is rushing to deploy.
aauth
The Git tag go get couldn't find: managing tags in a Go multi-module monorepo
I tagged a Go module v1.0.0, pushed it, and watched a teammate’s go get come back with invalid version: unknown revision. The tag was sitting right there on GitHub. The commit e...
go
Zero-trust for local agents: what SPIFFE gets right, and the session identity gap it leaves open
Every Goose session I run that touches infrastructure is authenticated with my credentials. My kubeconfig, my AWS profile, my GitHub token sitting in the shell environment. When...
devsecops
What Open Plugins gets right, and what it still needs
Every MCP server I’ve added to my Goose setup has been its own small project. Some are a git clone and a config path. Some need pip install and the right virtual environment. A ...
opinion
Selected writing elsewhere
- Agentic AI architecture: systems that think LinkedIn · 2025
- NIST's post-quantum cryptography standards: what it means for platform teams LinkedIn Pulse · 2024
- Service mesh proxy models and NIST SP 800-233 LinkedIn · 2024