BackStage

Hermes: Autonomous AI Agent with Persistent Learning

An experienced ML platform engineer describes how Hermes Agent — an open-source, local-first autonomous agent framework — is architecturally different from prior AI assistants and better suited to platform engineering. Hermes implements a three-layer memory (short-, medium-, long-term Skill Documents), a self-improvement loop the author calls GEPA (published at ICLR 2026 as an Oral), local SQLite data residency, multiple terminal backends (including SSH and Docker), built-in cron scheduling, and broad messaging integrations. The author shows concrete uses within his NeuroScale Kubernetes-based inference platform (drift diagnosis, pre-merge policy validation, incident RCA automation), highlights practical limitations (shallow domain reasoning, per-instance memory that does not yet federate, approval workflow risks), and notes Hermes’ rapid adoption claims (MIT license, large GitHub traction).

Framework for Engineering Build-vs-Buy Decisions

This article presents a practical framework for engineering leaders to decide whether to build or buy software capabilities. It reframes the binary "build vs buy" question into a multi-option decision space (build, buy SaaS, buy+customise, open-source+host, partner/outsource) and proposes a four-question test: (1) Is this core to your product? (2) Does a mature market solution exist? (3) What is the true total cost of ownership? (4) What is the blast radius of getting it wrong? The author recommends a default-to-buy policy if the four questions don't resolve a decision within two weeks, and offers a simple 3-year cost model (multiply vendor quote by 1.4; multiply build estimate by 2.5). The piece includes a decision matrix, domain-specific guidance (CI/CD, observability, AI/ML, security), real anonymized case studies, and an annual review template for revisiting decisions.

Agentic AI Workflows for Platform Engineering

This is Part 1 of a technical series explaining how to build agentic AI workflows for platform engineering teams. The author argues that improving developer velocity requires encoding team standards into the workspace (not just better prompts): steering files, skills, and agent definitions that provide persistent, role-specific context for AI agents. The post outlines a layered workspace model (e.g., a .kiro/ directory) that injects non-negotiable rules into every AI interaction, describes specialised agents for tasks like infrastructure authoring and security review, and details tool integrations (ticket trackers, CI/CD, AWS). The assumed stack includes AWS (multi-account), Terraform, GitLab CI, and AWS Secrets Manager. The article provides immediate starter steps (create a steering file and AGENTS.md) and previews later parts covering detailed steering files and GitOps/Kubernetes tooling.