GPT models have become much better since the early days of ChatGPT, but one lesson from GenAI security has survived every generation: a prompt is not a security boundary.

Prompt injection remains one of OWASP’s major LLM risks, and agentic systems have increased the potential consequences because a manipulated model may now have tools, credentials and access to other systems.

That means securing an AI application cannot consist of writing an increasingly elaborate system prompt telling the model what it must never do.

Input and output validation, least-privilege permissions, secret management, logging, dependency controls and conventional application security still have to surround the model.

GPT-5.6 may be far more capable than the models developers were experimenting with a few years ago. The surrounding software still needs to assume that the model can make mistakes, misunderstand instructions or be manipulated.

https://mkdev.me/posts/don-t-let-cyber-risk-kill-your-genai-vibe-a-developer-s-guide

If networking still feels like a collection of acronyms, this one from the mkdev archives is worth revisiting.

It walks through switches, routers, DNS, DHCP, NAT, VPNs and more, with practical examples along the way.

Read the full article: https://mkdev.me/posts/how-networks-work-what-is-a-switch-router-dns-dhcp-nat-vpn-and-a-dozen-of-other-useful-things

In the 96th mkdev dispatch, Pablo talks about what technology companies can learn from Spain’s football philosophy: competing with ambition, structure, creativity, and integrity to build systems that win for years instead of relying on shortcuts for short-term victories. Sign up for mkdev dispatch today!

https://mkdev.me/posts/you-don-t-win-by-cheating-96

Moving data into AWS is usually the easy part. Moving it around — and eventually back out — is where the pricing becomes more interesting.

AWS currently gives customers 100 GB of data transfer out to the internet per month for free, aggregated across supported AWS services and Regions, excluding China and GovCloud. CloudFront also includes 1 TB of free data transfer out each month. Beyond that, outbound pricing varies by Region and volume.

Inside AWS, topology matters too. Crossing Availability Zones can create transfer charges, and moving data between Regions has pricing based on the source and destination Regions. A seemingly small architectural choice can therefore become significant once you are moving terabytes every day.

Caching, compression, CloudFront, VPC endpoints and simply understanding your application's traffic patterns can all make a difference.

We break down these costs and optimization options in our guide to AWS data transfer pricing.

https://mkdev.me/posts/understanding-aws-data-transfer-costs

Logs, metrics and traces are only part of observability.

The real goal is understanding what happened, tracing problems across systems and making sure the right people get the right alerts — without alert fatigue.

Take a look at mkdev’s approach and arrange a call with us: https://mkdev.me/b/consulting/observability

AI explainability is easy to underestimate when AI is used for small, everyday tasks. If a model recommends the wrong article, produces a weak summary, or gives a slightly strange answer, the consequences are usually limited. We may be annoyed, but we can move on.

The problem begins when AI becomes part of decisions that people cannot simply ignore. A loan application, an insurance claim, a medical recommendation, a hiring process, a court case, or an autonomous system failure all create a very different expectation. In those situations, people need more than an output. They need a way to understand what influenced it, whether it was fair, and whether it can be challenged.

That is why AI explainability should not be treated as decoration around a model. A nice paragraph next to a prediction may improve the interface, but it does not automatically create accountability. The explanation has to match the decision, the risk, and the person who needs to use it.

As regulation catches up with AI adoption, companies will have to think about explainability much earlier in the product lifecycle. Not after deployment, not only when lawyers ask for it, and not as a checkbox. It has to be part of how AI systems are designed, tested, documented, and governed.

https://mkdev.me/posts/explaining-ai-explainability-vision-reality-and-regulation

Back in 2023, we tested AWS App Runner as a simpler way to deploy containers without managing all the usual infrastructure around ECS.

In 2026, App Runner is closed to new customers and AWS recommends ECS Express Mode instead. Watch the video to see where the idea worked — and where it didn’t: https://www.youtube.com/watch?v=E6E6HtrLs98

If you’re one of our Spanish-speaking subscribers and want to dive into containers, we have a free video course for you!

It covers Docker, docker-compose, Docker Swarm, Podman, Buildah, Firecracker and more — all in Spanish and completely free: https://www.youtube.com/playlist?list=PLNXwhzx0-DmRlCz9lKPLRWdv5visGlNNT

Cloud lock-in isn’t only about proprietary APIs and managed services.

Economics can create lock-in too.

Historically, one of the clearest examples was data egress: getting data into a cloud was often cheap or free, while getting large amounts of it back out could become expensive.

In 2026, the picture is changing. AWS, Google Cloud and Azure all offer programs that waive eligible egress charges when customers completely migrate away. In the EU, the Data Act has already started changing the rules around cloud switching, with switching charges due to be fully prohibited from January 2027.

But there is an important distinction: making it cheaper to leave a cloud does not mean data transfer itself has become free.

Applications still generate network costs between zones, regions, services and the public internet. Those costs can influence architecture just as much as compute or storage pricing.

The cloud is becoming easier to leave. Understanding the cost of moving data while you are still there remains just as important.

https://mkdev.me/posts/the-biggest-cloud-scam

Misconfigured RBAC, weak network policies or poorly managed secrets can leave a Kubernetes cluster exposed.

Our In-Depth Kubernetes Security Audit helps uncover these risks and gives your team a practical path to address them. Explore the audit and talk to us about your setup: https://mkdev.me/b/audits/kubernetes-security-audit

There are really two different problems hiding behind the term “AI explainability.”

The first is understanding how a model behaves in general. Global explainability methods can tell us which features tend to matter across a population and are particularly useful for developers who want to understand or debug a model.

The second is explaining one particular decision. Why was this loan application rejected? Why did this model produce this prediction? Local explainability methods such as LIME and SHAP try to answer those questions by building simpler approximations around individual cases.

The distinction matters because a population-level explanation doesn't necessarily tell you why something happened to one person. And a local approximation, however useful, isn't the same thing as opening up the original black box.

Explainability therefore isn't one technology solving one problem. It's a collection of approaches with different strengths, limitations, audiences and purposes.

We explored these questions in our article on AI explainability, and the distinction remains an important one for businesses working with increasingly complex AI systems: https://mkdev.me/posts/explaining-ai-explainability-the-current-reality-for-businesses

If Linux networking still feels like a collection of mysterious interfaces and commands, this one is worth revisiting.

Learn how teaming, Linux Bridge, tap interfaces and Traffic Control work together for fault tolerance and bandwidth management.

Read more: https://mkdev.me/posts/how-networks-work-part-two-teaming-for-fault-tolerance-bandwidth-management-with-traffic-control-tap-interfaces-and-linux-bridge

In the 97th mkdev dispatch Kirill explains the role Terraform has in this new age of AI agents. Also inside: AWS fast networking, Aurora DSQL pricing and more!

https://mkdev.me/posts/terraform-in-the-ai-agents-age-97

Imagine Google Cloud tells you you're spending roughly €500 a year on something you don't remember creating.

The obvious next step is to open FinOps Hub. You can inspect recommendations, look for potential savings and check where the spending is coming from.

But then you discover that the cost isn't an application server at all. It's infrastructure created to provide VPC connectivity for Cloud Run.

That's where the interesting part of FinOps starts.

Today, Direct VPC egress is Google's recommended approach for many Cloud Run workloads and avoids the compute cost of running Serverless VPC Access connector instances. It's a small architectural change that can remove an entire category of unnecessary spending.

We walk through this example, along with FinOps Hub, CUDs, cost allocation and billing analysis, in our Google Cloud FinOps article.

https://mkdev.me/posts/gcp-finops-hub-the-key-to-mastering-your-finances-on-google-cloud

CI/CD, infrastructure as code, Kubernetes and observability are powerful building blocks. Platform Engineering is about turning them into a coherent experience that helps teams ship efficiently at scale.

Read about our approach to Platform Engineering and arrange a call to discuss your setup: https://mkdev.me/b/consulting/platform-engineering

The interesting part about modern image models might not be image quality anymore.

It’s iteration.

Generating one impressive image is easy. Generating 50 or 100 versions, remembering what you disliked about previous attempts, making targeted corrections and gradually converging on something useful is a different problem.

That’s why combining Claude Code with image-generation models turned out to be more interesting than simply using another image-generation UI.

Claude can maintain the context of the task and use Nano Banana Pro or GPT Image 2 as tools. It can look at the output, notice that an object is positioned strangely or that the result doesn’t quite satisfy the request, and take another shot.

That turns image generation from a sequence of isolated prompts into something closer to an iterative creative workflow.

We explored the approach in this article and made the skills public:
https://mkdev.me/posts/unlimited-image-generation-with-nano-banana-pro-gpt-image-2-and-claude-code-skills

A useful Cloud Run distinction:

Services → request-driven applications
Jobs → run-to-completion workloads

And when the workload can be divided, Cloud Run Jobs can execute multiple tasks in parallel.

We break down the idea with a practical example here: https://www.youtube.com/watch?v=n8GyTp-kP_M

Want to understand containers beyond Docker? Our free Dockerless course takes you through OCI, open container standards, and the fundamentals that make modern containers work.

Article series: https://mkdev.me/posts/what-s-wrong-with-docker-introduction-to-the-dockerless-course
Video: https://www.youtube.com/playlist?list=PLozcbFx8FoPH30kYPbPuPsvxASWoLo9XB

For a while, microservices felt like the inevitable destination of every serious application.

Then teams discovered the other side of the equation: more services also mean more network calls, more APIs, more deployments, more dependencies, and more opportunities for things to fail in ways that are difficult to reproduce locally.

Google’s Service Weaver experimented with an interesting alternative. Developers could structure a Go application as a set of components without immediately committing every component to a separate service. Deployment topology could be decided later.

Service Weaver didn’t become the future of application development, and the project has since been archived. But the problem it was trying to solve hasn’t disappeared.

The interesting part of Service Weaver in 2026 isn’t the framework itself. It’s the question it leaves behind: should our code architecture really be so tightly coupled to our deployment architecture?

https://mkdev.me/posts/service-weaver-monolithic-or-microservice

Using Google Cloud doesn't automatically mean you're using the right services in the right way.

Our GCP Audit helps uncover unnecessary costs, security gaps, reliability issues and opportunities to simplify your infrastructure.

Take a look at what we cover, and schedule a conversation with our team: https://mkdev.me/b/audits/google-cloud-platform

There are two GenAI security concepts that are easy to mix up: jailbreaks and prompt injection.

A jailbreak generally tries to make an AI system ignore its restrictions and produce something it shouldn’t.

Prompt injection can go further. The goal may be to manipulate an AI system into taking actions or accessing systems and data that the attacker should never be able to reach.

As GenAI gets connected to more tools, APIs, and business systems, that distinction becomes increasingly important. The potential consequences move from “the chatbot said something bad” to data leakage, compromised systems, and operational disruption.

Our guide explains these risks from a product manager’s perspective, alongside another major issue: how employees and users handle sensitive data with GenAI tools.

Read the full article on mkdev: https://mkdev.me/posts/genai-security-risks-for-product-managers-dd73bdc2-4f2e-4227-93b3-375da081d906