The work,
written down.

Practitioner-level posts on AI infrastructure, agentic ops, retrieval, evals, cloud, and the older tech that shaped how we think about all of it. Published on Cookies Cookie Settings Echoes of the Machine. A curated entry-point index below.

Read everything on EOTM →

AI Infrastructure & Agentic Systems

Building production AI.

RAG, evals, prompts-as-code, the autonomy ladder, the AWS-native shape. The work behind the AI infrastructure practice.

The eval harness: how you know it's working before customers tell you

Test sets, golden examples, regression detection. How you find out the AI broke before a customer does.

What changes between Tier 2 and Tier 3 in practice

The promotion from execute-with-confirmation to bounded autonomy is the steepest cliff in the autonomy ladder.

Prompts as code: versioning, A/B, rollback

Git, CI, feature flags. The discipline the most-edited surface in an AI product needs.

Triage, diagnose, resolve: the three loops of an AI product

Different cost profiles, different SLAs. How to think about an AI product as a system, not a model call.

Retrieval is the secret-sauce surface: getting RAG right

Embeddings, chunking, hybrid search, the prompt template that pulls in the right context.

The AWS-native shape I actually start with

API Gateway, Lambda, Cognito, RDS+pgvector, Bedrock, S3, EventBridge, SQS, CloudFront, CloudWatch.

The infrastructure underneath.

Data layer, auth, multi-tenancy, the cloud/local hybrid split. The work behind the cloud practice.

Personal AI & Local-First

What runs on the desk.

Apple Silicon, Ollama, mflux, the Mac Studio side of the hybrid stack. The math on why it pays for itself.

The Mac Studio side of the stack

mflux, ollama/mlx-lm, fine-tuning, whisper, a batch runner. The math on why it pays for itself in months.

All Personal AI posts on EOTM

The full archive of local-first AI writing: workstation builds, model choices, and the practical-vs-cloud tradeoffs.

All Automation posts on EOTM

Older and newer automation work: IaC patterns, agentic ops, and the operational glue underneath both.

Yesterday's tech, today's instincts.

Echoes of the Machine spends a lot of time on the older work (VMware, vRA, NSX, OneFuse, the 386 era) because the patterns hold. The names change. The shape doesn't.

Want a post that goes deeper on a topic?

Most engagements start with a conversation about something we've written. Tell us which one resonated.

Start a conversation →