Hey {{first name | there}}. How we work with technology is changing, and so are how we prove what we know, build with AI, and keep the tools we rely on secure. This week's developments offer a few interesting examples of where those changes are taking us.
In today’s roundup:
Microsoft introduces skills-based credentials.
LinkedIn traces the member journey with OpenTelemetry.
Claude Code restrictions frustrate developers.
Airbnb plans an AI travel agent for 2027.
Researchers uncover a bypass for ChatGPT Lockdown Mode.
LATEST DEVELOPMENT
🧠PRESENTED BY EVERYTHINGDEVOPS
Kubernetes can be a lot to take in when you're just starting. The best way to understand it is to get a cluster running and start working with it.
In this hands-on crash course, you'll build a local cluster, deploy your first application, scale it, expose it in a browser, and deliberately break things to see how Kubernetes responds.
You'll also learn the core concepts behind pods, deployments, services, and YAML configurations.
By the end, you'll have deployed an application yourself and a foundation to build on.
Image credit: TheRegister
The Rollout: Microsoft has introduced Pro Badges, a new credential that measures how people use its products in everyday work rather than testing them through traditional exams. The first badge recognises practical proficiency with GitHub Copilot.
The details:
Users earn the badge by signing in to GitHub's learning platform, where Microsoft's proficiency engine analyses their product usage.
Five proficiency levels range from beginner to expert. Users must reach the proficient level to qualify.
Assessment considers sustained hands-on use across multiple repositories, including model selection, customisation, and everyday development workflows.
Why this matters: Traditional certifications demonstrate knowledge under exam conditions, but they don't always show how someone applies those skills at work. Microsoft's approach offers another way to demonstrate practical experience, particularly for developers who use its tools regularly.
The Rollout: LinkedIn has built an OpenTelemetry-based observability platform that traces member interactions across its iOS, Android, and web applications, connecting frontend activity to the backend services and databases processing each request.
The details:
A new identifier, pageForestId, groups related requests from the same page, complementing OpenTelemetry's existing traceId.
Telemetry flows through Apache Kafka into ClickHouse, with coordinated sampling to preserve related client and backend traces.
Custom instrumentation handles mobile lifecycle challenges, while data controls and redaction help prevent sensitive member information from entering traces.
Why this matters: Traditional distributed tracing follows individual requests, but a user's experience can involve dozens of related requests. Connecting the full journey helps engineers move from a reported problem to its underlying technical cause faster.
Image credit: TheNewStack
The Rollout: Anthropic has added restrictions to Claude Code to prevent users from recreating existing video games. However, developers report that Claude Opus 5.5 is also refusing legitimate reverse-engineering and software-analysis tasks.
The details:
Anthropic confirmed it added two instructions to Claude Code's prompt to prevent game cloning that could infringe intellectual property.
Developers report stalled client projects and refusals when analysing software. Some have switched to older Claude models to continue working.
Anthropic says original game development remains allowed, but has not clarified whether the broader reverse-engineering refusals stem from separate security restrictions.
Why this matters: Hosted coding agents can change behaviour when providers update their instructions or safeguards. Developers may suddenly lose access to workflows they depend on, even without changing their code or prompts. This raises questions about reliability, transparency, and how much control developers have over the tools they use.
💡TOGETHER WITH EVERYTHINGDEVOPS
What happens when you bring platform engineers, DevOps professionals, and cloud-native experts into one room?
We spent the day at Amazon’s London HQ for the Day of Containers, with talks from AWS, Checkout.com and Spacelift, followed by a hands-on workshop where attendees got to build with AWS and Spacelift.
Divine also caught up with speakers and attendees to hear their perspectives on EKS, policy-as-code, platform engineering, and what teams are working on in practice. There was plenty of networking, a little cloud-native trivia, and a few surprises along the way.
Take a look at what happened at the event.
Image credit: Airbnb
The Rollout: Airbnb plans to launch its AI travel agent in the first half of 2027. Rather than relying on a text-based chatbot, the agent will create a personalised visual interface using travel-specific tools such as maps, photos, comparisons, calendars, and itineraries.
The details:
The interface will adapt to each traveller's needs, helping them explore destinations, compare accommodation, and plan trips.
The agent will also handle post-booking support, including reservation changes, payment issues, refunds, rebooking, and 24-hour safety assistance.
Airbnb plans to expand its capabilities beyond accommodation to include travel services and experiences.
Why this matters: Airbnb is betting that travel planning needs more than a general-purpose AI assistant. Its approach puts visual discovery and decision-making alongside practical support, showing how companies could build specialised AI experiences around the services they already provide.
Image credit: TheHacker News
The Rollout: Security researchers at Check Point disclosed a now-patched vulnerability that could bypass ChatGPT's Lockdown Mode protections. The flaw exploited how HTML previews handled pop-up windows, creating a route around the sandbox restrictions designed to prevent data from leaving a conversation.
The details:
Attackers could use crafted prompts, malicious GPTs, or shared chats to steal conversation data, deliver malware, or launch phishing attacks.
In the desktop application, the vulnerability could enable zero-click data exfiltration, meaning sensitive chat data could be sent out without the user taking an action.
OpenAI addressed the vulnerability on September 18, 2026, following responsible disclosure by Check Point.
Why this matters: Lockdown Mode was designed to reduce the risk of prompt injection and unauthorised data transfers. This finding shows how weaknesses in surrounding application features can undermine AI security controls, even when the model itself is not the source of the vulnerability.
QUICK LINKS
📅EVENTS
🇪🇺 🇬🇧 🇩🇪 Europe
Title | Date | Location |
November 9 | Berlin (Germany) | |
November 11 -12 | London (UK) |
🇺🇸🇨🇦North America
Title | Date | Location |
November 8 | Salt Lake City, UT | |
November 9 | Salt Lake City |
🇳🇬 🇰🇪 Africa
Title | Date | Location |
November 13-14 | Ikorodu, Nigeria | |
November 13-14 | Nairobi, Kenya | |
November 13-14 | Banjul, Gambia |
🌎Asia
Title | Date | Location |
November 14-15 | Bengaluru, India | |
November 14-15 | Bangkok, Thailand |