TLDR DevOps 2026-01-09

Lloyds Banking Group has been around since 1765. Now they're running one of the most ambitious digital transformations in UK banking — with feature flags at the center of it.

In this live webinar, Daniel Woollery (Head of Digital Services, Lloyds) shares how the bank scaled FeatureOps from scattered team experiments to an enterprise-wide platform.

📋 On the agenda: migrating hundreds of flags off legacy systems, building governance that accelerates delivery, and rolling out GenAI with feature-level risk controls.

🕰️ Webinar starts January 19 @ 10AM ET. Save your spot

Red Hat expanded its collaboration with NVIDIA to deliver Day 0 support for the NVIDIA Rubin platform across Red Hat Enterprise Linux, OpenShift, and Red Hat AI. The partnership targets rack-scale enterprise AI with optimized performance, security, and hybrid cloud consistency.

Apache Spark 4.1 introduces Spark Declarative Pipelines (SDP) for a declarative data engineering approach, and enables Real-Time Mode in Structured Streaming to achieve single-digit millisecond latencies for stateless workloads. The update also delivers GA support for SQL Scripting, expands PySpark capabilities with Arrow integration, and enhances Spark Connect for machine learning.

Kubernetes v1.35 introduces a beta opt-in solution for CSI drivers, enabling them to securely receive service account tokens through the `secrets` field in `NodePublishVolumeRequest` instead of the less secure `volume_context`. This addresses past logging vulnerabilities, such as CVE-2023-2878, by providing a safer, more consistent method for handling sensitive data.

Modern software development prioritizes rapid feature delivery and short-term impact metrics over long-term code quality and maintainability. Incentive structures, Agile backlog pressure, low deployment risk, and a lack of emphasis on craftsmanship have collectively contributed to declining code quality across organizations.

Unkey migrated from Cloudflare Workers to stateful Go servers, achieving 6x lower latency by eliminating external cache network requests and simplifying architecture from complex distributed pipelines to straightforward in-memory batching. The move enabled self-hosting and platform independence, improved developer experience, and reduced operational complexity.

AI-native engineering = engineers set the goals and agents perform the work. When agents handle log analysis, metric correlation, and timeline reconstruction during an incident, engineers can shift their focus to architectural decisions and system design rather than tactical investigation.

Download the free AI for Prod ebook to learn more about how AI is changing core engineering workflows.

OpenSpec is a spec-driven development (SDD) framework for AI coding assistants that aims to align humans and AI by locking intent with a lightweight specification workflow before any code is written. The system provides deterministic, reviewable outputs, addressing the unpredictability of AI when requirements are confined to chat history.

Next AI Draw.io is an AI-powered Next.js web application that enables users to create, modify, and enhance draw.io diagrams using natural language commands and AI-assisted visualization. This tool supports various diagram types and recommends powerful models like Claude Sonnet 4.5 for optimal performance.

A routing leak in Venezuela on January 2, involving AS8048 (CANTV), was noted by a cybersecurity newsletter amid news of Nicolás Maduro's capture, initially suggesting potential malfeasance. However, Cloudflare Radar data showed eleven similar, likely accidental, Type 1 hairpin route leaks by CANTV since early December, indicating the issue stemmed from insufficient routing export and import policies rather than malicious intent.

Grafana Cloud's Service Center provides a unified hub for monitoring service reliability, performance trends, incidents, SLOs, alerts, and on-call rotations to prevent engineer burnout and improve operational culture. The feature centralizes scattered data into service pages that help teams make data-driven decisions about reliability improvements and workload balancing.

PostgreSQL 18 significantly enhances the RETURNING clause by introducing OLD and NEW aliases across INSERT, UPDATE, DELETE, and MERGE, enabling developers to access before-and-after row values in a single atomic statement. This removes the need for triggers or extra queries, simplifies change tracking and auditing, and makes MERGE-based upserts far more powerful and maintainable.

GitHub Actions automates CI/CD pipelines for deploying microservices to Azure Kubernetes Service by building Docker images, pushing them to Azure Container Registry, and deploying to AKS clusters using kubectl.

HolmesGPT, an open-source AI troubleshooting agent for Kubernetes and cloud-native environments, was accepted as a CNCF Sandbox project in October.