arnab@platform:~$/posts --latest
[ 02 ]/posts// 12 entries · ordered by date
Technical articles on AI, Kubernetes, Platform Engineering, and Cloud Infrastructure.
20268 posts
┌─mirador-headless-mac-agents.mdx
9 min read
The Last Mile of Headless Mac Agents Is the Screen
Jun 21, 2026
The Last Mile of Headless Mac Agents Is the Screen
SSH solved most remote agent work. Mirador exists for the moments when the Mac still needs a real screen.
aiagentsmacosios
┌─agent-observability-starts-in-the-runtime.mdx
11 min read
Causality Is the Contract Between Runtime and Observability
May 16, 2026
Causality Is the Contract Between Runtime and Observability
Observability can show activity, but durable causality depends on the runtime signals it receives.
aiagentsobservabilityopenclaw
┌─why-llm-routing-needs-an-explanation-layer.mdx
9 min read
Why LLM Routing Needs an Explanation Layer
Apr 21, 2026
Why LLM Routing Needs an Explanation Layer
Routing between different models is useful. Being able to see why a model was chosen is the part that changes how you build agents.
aiagentsroutinganthropic
┌─agentweave-observability.mdx
12 min read
AgentWeave: Observability for Multi-Agent Systems
Mar 21, 2026
AgentWeave: Observability for Multi-Agent Systems
When you're running agents that delegate to other agents, logs don't show you the chain. Here's how I built observability into my home lab setup using OpenTelemetry and W3C PROV-O.
aiagentsobservabilityopentelemetry
┌─a2a-self-hosted-multi-agent-architecture.mdx
14 min read
A2A: Building a Self-Hosted Multi-Agent System
Mar 6, 2026
A2A: Building a Self-Hosted Multi-Agent System
Most multi-agent setups fail because agents are too generic. Here's what happens when you strip each one down to exactly what it needs and connect them with a real protocol.
aiagentsarchitecturehomelab
┌─nixclaw-voice-vision-agent.mdx
12 min read
Building NixClaw: Giving My AI Agent Eyes and Ears
Feb 25, 2026
Building NixClaw: Giving My AI Agent Eyes and Ears
From a Siri Shortcut to a full voice + vision AI that can see through smart glasses and take actions on my behalf. The journey of building NixClaw, and why the 'last mile' of AI assistants is harder than it looks.
aimultimodalagentsvoice
┌─shipping-on-the-go.mdx
10 min read
Shipping on the Go: Something Shifted
Feb 17, 2026
Shipping on the Go: Something Shifted
Something shifted in the last few weeks. Not just better AI tools, but a fundamental change in how I build. Thought to execution, no barrier. Here's what that looks like.
aiproductivityengineeringworkflow
┌─gpu-on-demand-wake-on-lan.mdx
8 min read
GPU-on-Demand: Wake-on-LAN for ML Workloads
Feb 13, 2026
GPU-on-Demand: Wake-on-LAN for ML Workloads
How I set up on-demand GPU access for ML workloads without leaving a power-hungry PC running 24/7. Save power, reduce heat, and still get fast embeddings when you need them.
mlgpuhomelabautomation