<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <title>Platform Monkey</title>
  <subtitle>Platform engineering &amp; AI tool reviews: Kubernetes, CNCF, Terraform, OpenTofu, Argo CD, Flux, GitOps, Helm, Crossplane, Backstage, internal developer platforms (IDP), observability, and self-service.</subtitle>
  <link href="https://platform-monkey.com/feed.xml" rel="self"/>
  <link href="https://platform-monkey.com/"/>
  <updated>2026-09-19T00:00:00Z</updated>
  <id>https://platform-monkey.com/</id>
  <author>
    <name>Yuval Glasman</name>
  </author>
  
  <entry>
    <title>Claude Code Adopts AGENTS.md: The Last Coding-Agent Holdout Blinks</title>
    <link href="https://platform-monkey.com/articles/claude-code-agents-md-adoption"/>
    <updated>2026-09-19T00:00:00Z</updated>
    <id>https://platform-monkey.com/articles/claude-code-agents-md-adoption</id>
    <summary>Claude Code 2.1.277 reads AGENTS.md — the last major holdout joins the instruction-file standard. Precedence rules, Bedrock gaps, and what platform teams must govern.</summary>
  </entry>
  
  <entry>
    <title>Cloudflare&#39;s 100TB Hash-Ring Fix: The Cost of &#39;Just Add More Hashes&#39;</title>
    <link href="https://platform-monkey.com/articles/cloudflare-ketama-100tb-ram"/>
    <updated>2026-09-19T00:00:00Z</updated>
    <id>https://platform-monkey.com/articles/cloudflare-ketama-100tb-ram</id>
    <summary>Cloudflare cut 100TB of RAM by deleting 90% of its consistent-hash points. The math every load-balancer owner should run before scaling out.</summary>
  </entry>
  
  <entry>
    <title>Cloudflare Quick Tunnels for Coding Agents: The Viral Pitch vs. What Actually Ships</title>
    <link href="https://platform-monkey.com/articles/cloudflare-quick-tunnels-agents"/>
    <updated>2026-09-19T00:00:00Z</updated>
    <id>https://platform-monkey.com/articles/cloudflare-quick-tunnels-agents</id>
    <summary>We ran Cloudflare&#39;s viral Quick Tunnels with --output json. Real agent ergonomics, the 200-request cap, edge TLS, and the egress policy your team needs.</summary>
  </entry>
  
  <entry>
    <title>AI Code Review Agents in 2026: What Works, What Annoys Seniors, and What It Costs</title>
    <link href="https://platform-monkey.com/guides/ai-code-review-agents"/>
    <updated>2026-09-19T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/ai-code-review-agents</id>
    <summary>Copilot code review added approvals, auto-resolution and ensemble agents in three weeks. How AI reviewers behave in production - and what they cost.</summary>
  </entry>
  
  <entry>
    <title>Kubernetes Cost Visibility in 2026: OpenCost, VPA InPlace, and the Right-Sizing Math That Actually Saves Money</title>
    <link href="https://platform-monkey.com/guides/k8s-cost-visibility-right-sizing"/>
    <updated>2026-09-19T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/k8s-cost-visibility-right-sizing</id>
    <summary>Kubernetes cost visibility without guesswork: OpenCost 1.121 allocation math, VPA InPlaceOrRecreate right-sizing, Goldilocks namespaces, and the unit economics of requests.</summary>
  </entry>
  
  <entry>
    <title>LLM Routers and Gateways After Jev: LiteLLM, OpenRouter, and the Decision-Only Workload Split</title>
    <link href="https://platform-monkey.com/guides/llm-router-gateways"/>
    <updated>2026-09-19T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/llm-router-gateways</id>
    <summary>TypeSafe&#39;s Jev reads choice logits without decoding a single token. Here is the gateway architecture that should route decision traffic differently from generation traffic.</summary>
  </entry>
  
  <entry>
    <title>Loki vs. Elasticsearch in 2026: Object-Store Log Economics Measured on Real Disk</title>
    <link href="https://platform-monkey.com/guides/loki-vs-elasticsearch"/>
    <updated>2026-09-19T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/loki-vs-elasticsearch</id>
    <summary>We ran the same 300k-line corpus through Loki 3.7.8 and Elasticsearch 9.5.4. The storage gap is real, but the new logsdb_columnar mode changes the math.</summary>
  </entry>
  
  <entry>
    <title>Bonsai 2 27B: Ternary Weights Meet the 8 GB GPU — and the Upstream Tax</title>
    <link href="https://platform-monkey.com/articles/bonsai-2-27b-ternary-llm"/>
    <updated>2026-09-18T00:00:00Z</updated>
    <id>https://platform-monkey.com/articles/bonsai-2-27b-ternary-llm</id>
    <summary>PrismML&#39;s ternary Qwen3.8-27B runs in 5.9 GB at 98.2% retention. Real numbers, the llama.cpp fork tax, and who should skip it.</summary>
  </entry>
  
  <entry>
    <title>ZCode&#39;s Silent .git Upload: Why AI Coding Harnesses Are Now a Data-Perimeter Problem</title>
    <link href="https://platform-monkey.com/articles/zcode-git-history-exfiltration"/>
    <updated>2026-09-18T00:00:00Z</updated>
    <id>https://platform-monkey.com/articles/zcode-git-history-exfiltration</id>
    <summary>Z.ai&#39;s ZCode packages whole workspaces — 86.6% of the payload is .git — and uploads them to Aliyun OSS under server-held keys. Detection and containment.</summary>
  </entry>
  
  <entry>
    <title>Running AI Agents on Kubernetes: Sandboxes, Egress, and the Isolation Lie</title>
    <link href="https://platform-monkey.com/guides/ai-agents-on-kubernetes"/>
    <updated>2026-09-18T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/ai-agents-on-kubernetes</id>
    <summary>A new 176-setting study proves harness design, not the model, decides coding-agent outcomes. This is the Kubernetes architecture it implies.</summary>
  </entry>
  
  <entry>
    <title>Cilium Service Mesh: The Sidecar-Free eBPF Mesh on Kubernetes 1.37</title>
    <link href="https://platform-monkey.com/guides/cilium-ebpf-service-mesh"/>
    <updated>2026-09-18T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/cilium-ebpf-service-mesh</id>
    <summary>A production deep-dive into Cilium&#39;s sidecar-free service mesh: eBPF datapath, ztunnel mTLS, Gateway API, L7 policy, Hubble, and when to skip it.</summary>
  </entry>
  
  <entry>
    <title>Gateway API vs. Ingress: The Migration Every Kubernetes Team Now Owes</title>
    <link href="https://platform-monkey.com/guides/gateway-api-vs-ingress"/>
    <updated>2026-09-18T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/gateway-api-vs-ingress</id>
    <summary>ingress-nginx retired in March 2026 and Gateway API v1.6 is GA. The real migration guide: CRD traps, controllers, annotation conversion, and cutover.</summary>
  </entry>
  
  <entry>
    <title>GitOps Secrets After the OpenAI Breach: SOPS, External Secrets, and the Vault Question</title>
    <link href="https://platform-monkey.com/guides/gitops-secrets-management"/>
    <updated>2026-09-18T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/gitops-secrets-management</id>
    <summary>How to run secrets in GitOps without becoming the next breach: SOPS+age, Sealed Secrets, External Secrets Operator, Vault trade-offs, and the identity controls that matter.</summary>
  </entry>
  
  <entry>
    <title>OpenTofu vs. Terraform: The 2026 Fork Decision Matrix</title>
    <link href="https://platform-monkey.com/guides/opentofu-vs-terraform"/>
    <updated>2026-09-18T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/opentofu-vs-terraform</id>
    <summary>OpenTofu 1.13 cuts WinRM and adds Symbol Libraries while Terraform 1.17 bets on Policy GA. The real fork decision matrix for platform teams, Sept 2026.</summary>
  </entry>
  
  <entry>
    <title>The Backstage Trap: The TCO of &quot;Single Pane of Glass&quot; Portals</title>
    <link href="https://platform-monkey.com/articles/backstage-trap"/>
    <updated>2026-09-17T00:00:00Z</updated>
    <id>https://platform-monkey.com/articles/backstage-trap</id>
    <summary>A technical analysis of the maintenance overhead and plugin dependency hell associated with self-hosting Backstage.</summary>
  </entry>
  
  <entry>
    <title>The Artifactory Message Board: A Case Study in Agentic Side-Channels</title>
    <link href="https://platform-monkey.com/articles/openai-huggingface-incident"/>
    <updated>2026-09-17T00:00:00Z</updated>
    <id>https://platform-monkey.com/articles/openai-huggingface-incident</id>
    <summary>Deep-dive into the July 2026 incident where OpenAI&#39;s internal research models bypassed sandboxes via shared package management infrastructure.</summary>
  </entry>
  
  <entry>
    <title>The Shift Left Lie: Infrastructure Dumping and Developer Poverty</title>
    <link href="https://platform-monkey.com/articles/the-shift-left-lie"/>
    <updated>2026-09-17T00:00:00Z</updated>
    <id>https://platform-monkey.com/articles/the-shift-left-lie</id>
    <summary>An analysis of how &quot;Shift Left&quot; became infrastructure dumping on developers, increasing cognitive load instead of empowering teams with paved roads.</summary>
  </entry>
  
  <entry>
    <title>ML-Driven Query Planning: 81% Faster or Just Unstable?</title>
    <link href="https://platform-monkey.com/guides/ai-postgres-query-plans"/>
    <updated>2026-09-17T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/ai-postgres-query-plans</id>
    <summary>Analysis of using 4B parameter models to replace heuristic-based Postgres query planners.</summary>
  </entry>
  
  <entry>
    <title>ArgoCD vs. Flux: Reconciliation Loops, State Drift, and the &quot;Sync Loop of Death&quot;</title>
    <link href="https://platform-monkey.com/guides/argocd-vs-flux"/>
    <updated>2026-09-17T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/argocd-vs-flux</id>
    <summary>A technical deep-dive into ArgoCD and Flux reconciliation engines, covering sync loops, HPA conflict resolution, and multi-cluster state management.</summary>
  </entry>
  
  <entry>
    <title>AWS vs. Azure vs. GCP: A Platform Engineer&#39;s Cold-Blooded Comparison</title>
    <link href="https://platform-monkey.com/guides/aws-vs-azure-vs-gcp"/>
    <updated>2026-09-17T00:00:00Z</updated>
    <id>https://platform-monkey.com/guides/aws-vs-azure-vs-gcp</id>
    <summary>Moving past the pricing sheets. We analyze the Big Three from a platform perspective: API ergonomics, K8s managed services, and the cost of lock-in.</summary>
  </entry>
  
</feed>
