Every week, AI helps your team work faster, but it also increases your data’s exposure. Files move between new tools, models use sensitive data, and traditional DLP often misses the most important context.

OnFebruary 3 at 11:00 AM PT, we’ll introduce Cyberhaven’s data lineage powered and unified DSPM and DLP platform. You’ll see how one AI-native solution can finally keep up with the way data really moves.

Join us live to see:

Don’t wait for another AI-related incident to reveal gaps in your data security. Reserve your spot and be among the first to see how a unified DSPM and DLP platform can change how your organization protects its most important data.

The official Kubernetes Dashboard is getting archived after a decade. No active contributors or maintainers left. End of an era for one of the earliest K8s UI projects.

Meanwhile, someone trained LLMs on three years of incident postmortems and built systems that predict outages 15-45 minutes before alerts fire. We're also covering K8s 1.35's in-place pod restarts, why learning Linux primitives makes Kubernetes finally click, and a Palo Alto DoS flaw that crashes firewalls into maintenance mode.

Plus: 20+ tools that auto-generate K8s diagrams and a game where you fix 50 broken clusters to learn.

Cheers,

Shreyans Singh

Editor-in-Chief

Use code FINAL30

Kubernetes 1.35 lets you restart entire pods in-place

K8s 1.35 adds in-place pod restart (alpha, behind RestartAllContainersOnContainerExits gate) which is huge for AI/ML workloads. Previously if an init container corrupted the environment or a sidecar failed, you had to delete the entire pod and let the scheduler recreate it: slow and expensive. Now you can trigger a full restart that preserves pod UID, IP, network namespace, sandbox, volumes, everything except ephemeral containers. All init containers rerun from scratch, giving you a clean state.

Training AI on your incident history predicts outages 15-45 minutes early

Someone trained LLMs on three years of incident postmortems and built systems that predict failures 15-45 minutes before traditional alerts fire.

The trick is extracting causal embeddings. Not just "symptom and cause are related" but learning the transformation from "what we observed" to "what was actually wrong." They decompose incidents into structured reasoning chains, create separate vector spaces for symptoms/causes/resolutions/precursors, then continuously pattern-match current system state against historical precursor embeddings.

Every tool that generates Kubernetes architecture diagrams

Huge GitHub repo comparing 20+ tools that generate K8s architecture diagrams from manifests, APIs, Helm charts, etc.

KubeDiagrams leads with 47+ resource types supported, reads from manifests/kustomize/Helm/API, outputs to PNG/SVG/PDF/DOT, supports namespace/label clustering. Most tools use Python with Diagrams library, some use Go/TypeScript/Java. Common pattern: 60% support KIS (Kubernetes Icons Set), 45% do namespace clustering, 95% show Services, 80% show Deployments.

Learn Kubernetes by fixing 50 broken clusters

Open source game-based K8s training with 50 progressive challenges across 5 worlds (Core Basics, Deployments, Networking, Storage, Security). Each level breaks something in K8s and you fix it using kubectl. Has real-time monitoring with "check" command, progressive hints, step-by-step guides, post-mission debriefs explaining why your fix worked.

Palo Alto patched a DoS flaw that crashes firewalls into maintenance mode

Palo Alto patched CVE-2026-0227 (CVSS 7.7), a DoS vulnerability in PAN-OS firewalls with GlobalProtect enabled that lets unauthenticated attackers crash firewalls into maintenance mode. PoC code already exists and a researcher reported it, though no active exploitation yet. This is almost identical to CVE-2024-3393 from late 2024 which was a zero-day.

Use code EARLY40

Why you should learn Linux before diving into Kubernetes

Docker didn't invent containers. It wrapped existing Linux features (cgroups, namespaces) that Google had been using for years into a simple interface anyone could use. Every K8s feature relies on Linux primitives: pod isolation uses namespaces (PID, network, mount, user, IPC), resource limits use cgroups, networking uses iptables/nftables for ClusterIP services and NAT, network policies use packet filtering, images use OverlayFS for layered filesystems, Cilium uses eBPF for high-performance networking instead of iptables. When you create a Pod, you're orchestrating Linux isolation and resource management tools. Understanding namespaces, cgroups, network filtering makes K8s and Docker click—you realize they're just convenient wrappers over powerful Linux capabilities. Learn the foundation first, the abstractions make way more sense after.

Auto-comment K8s manifest changes on PRs

Go tool that receives GitHub webhooks for PRs, auto-discovers ArgoCD apps configured with that repo as source, generates diffs against live state using ArgoCD CLI, and comments on PRs with markdown showing what would change. No per-repo configuration needed.

How etcd actually works (and why Kubernetes uses it)

etcd is a strongly consistent distributed key-value store using the Raft consensus algorithm. All writes go through an elected leader, changes replicate to followers, new elections happen if leader dies. Production clusters typically run 3 or 5 nodes (odd numbers only since you need majority for availability). K8s stores everything under /registry prefix with naming like /registry/pods/<namespace>/<pod-name> , uses prefix queries and watch subscriptions for real-time updates. This is how controllers and operators subscribe to resource changes.

Kubernetes Dashboard is being archived after a decade

The official Kubernetes Dashboard project is getting archived after no active contributors and maintainers running out of time to work on it. Started in 2015 when K8s was still new, it served the community for over a decade but ecosystem needs have changed significantly. End of an era for one of the earliest K8s UI projects, but makes sense given how much the tooling landscape has evolved since 2015.

Self-healing infrastructure is running in production right now

Autonomous healing infrastructure isn't science fiction. It's operational in production serving millions of users, and the difference from past attempts is reasoning capability. The architecture needs four pieces: decision engine combining rule-based policies with LLM reasoning for edge cases, safety sandbox that never executes directly in prod (snapshots state, enhanced monitoring, automatic rollback on any degradation), graduated action library (green/yellow/red based on risk), and learning loop where every action generates training data to improve confidence scores.

📢 If your company is interested in reaching an audience of developers and, technical professionals, and decision makers, you may want toadvertise with us.

If you have any comments or feedback, just reply back to this email.

Thanks for reading and have a great day!