Last Week in AI: 4th Edition

Last Week in AI: 4th Edition

November 19, 2025

Hey all, and welcome back to Last Week in AI, your weekly roundup of what engineers, builders, and hiring leaders need to know.

This week’s updates focused on giving developers more control over how AI is integrated, shipped, and secured. Microsoft upgraded the core .NET toolchain. Postman and Solo.io launched new primitives for APIs and agents. Helm modernized packaging for cloud-native delivery. And VibeGuard introduced a real-time safeguard for AI coding assistants.

Here’s what shipped:

1 | Microsoft rolls out agent support in Visual Studio 2026 and .NET 10

At .NET Conf 2025, Microsoft announced Visual Studio 2026 and .NET 10, both focused on bringing agent orchestration into core workflows. Visual Studio now includes built-in AI agents that flag issues, suggest fixes, and assist in C# and C++ development. 

Article content
Source: The Register

.NET 10 adds an agent framework for orchestrating multi-step AI workflows directly inside ASP.NET Core apps, unifying prior tools like Semantic Kernel.

C# 14 also shipped with better syntax and memory handling, and both releases are fully backward compatible. This stack update gives developers a full agent layer from runtime to IDE. If you understand how to structure agent behavior at both levels, you will be able to ship faster and with clearer intent.

2 | Postman adds auto-SDK generation with liblab integration

Postman now supports automatic SDK generation, thanks to its acquisition of liblab. You can turn any OpenAPI spec into versioned SDKs across multiple languages, with updates synced automatically. This reduces the overhead of maintaining client wrappers across platforms.

Teams can enforce consistent structure without needing language-specific boilerplate, and ship SDKs with every release. It also simplifies internal API adoption — developers consuming APIs get a ready-to-use, versioned interface without needing to read the spec or handle low-level edge cases.

Article content

3 | Solo.io launches agentregistry: infra for managing AI Skills

Solo.io released agentregistry, an open-source registry for organizing and governing agent Skills. It includes self-contained folders with scripts, instructions, and task-specific logic that agents can load and execute at runtime. The registry supports versioning, dependency graphs, policy controls, and can be run locally or curated for team-wide access.

This offers a structured way to publish, manage, and compose agent behavior using familiar DevOps patterns. Skills can now be packaged like services, shared across teams, and updated independently of the agents that consume them. 

4 | Helm 4 ships with secure packaging and better deployment tooling

Helm 4 (the first major release in six years) brings chart signing, test automation, and a new plugin system with WebAssembly support. The update also improves Helm’s SDK, embeds CLI commands for use in other tools, and supports modern Kubernetes features like server-side apply.

If you’re working on multi-service apps or deploying AI workloads across clusters, Helm 4 makes packaging more secure, plugins more portable, and delivery workflows easier to integrate into CI/CD. 

Teams that build and ship reproducible systems will benefit most from the tighter controls and modern extension points.

5 | Epic Games launches Unreal Engine 5.7 with performance upgrades

Epic released Unreal Engine 5.7, improving performance and fidelity across real-time workflows. The update refines Nanite rendering, adds native support for volumetric path tracing, and introduces new tooling for physics, animation, and UI.

Article content
Image: Yahoo! Tech

For developers working at the intersection of simulation, spatial systems, and real-time interactivity, UE5.7 offers a faster path from code to embodied experience plus more control over how those environments are built and tested.

The through-line

The developer role is expanding across more of the build surface. Agent logic can be versioned and reused like libraries. SDKs are generated and maintained automatically. Infrastructure layers like Helm and Kubernetes offer more direct control over packaging and deployment. Simulation environments are emerging as programmable spaces for system behavior and interaction.

Each update adds a new point where structure and intent can be codified into the system itself. Now we have to understand where/how to define behavior, enforce boundaries, and design for reuse.

What’s new at HackerRank

We’ve partnered with byteXL to bring global hiring benchmarks to engineering campuses across India. See the announcement.

We’re also hosting a backend-focused coding challenge for engineers in Bangalore. If you care about writing clean, scalable systems, join us:

Challenge Date: Nov 23, 12–1 PM IST | Register by: Nov 20, 7 PM IST

Just Released & Coming Up

Next-generation developers are applying these tools with precision. They design workflows with intent, manage complexity at scale, and structure systems to be easier to extend and reason about.

See you in the next edition!

In case you missed it: Edition #1 (10/9) | Edition #2 (10/24) | Edition #3 (11/6)

Mercy Zheng

Software Development Director

15h

These updates highlight how AI, automation, and modern tooling are giving developers more control over tasks, APIs, and deployments. Powerful advancements for building smarter software.

MEGATORC PARAFUDADEIRA COM TORQUE CONTROLADO – 01 à 15.000 Nm. BATERIA - Motor sem escovas. - Muito Rápida, Confiável e Ergonômica. - Processo de aperto controlado por Torque e Ângulo. - Especificação e documentação do processo de aparafusamento possível via WLAN. - Visor LCD colorido, multilíngue, com mensagens de controle e aviso. - Alto desempenho de torque e precisão: aperto confiável de conexões de parafusos de até 15.000 Nm com repetibilidade de ±1%. - Os valores de torque podem ser predefinidos, valores específicos do cliente. - Opção: torques contínuos (etapas de 1 Nm). - Opção: Especificação e documentação do processo de aparafusamento via WLAN. - Posição Ergonômica: o acionamento e a caixa de engrenagens são desacoplados por meio de uma junta giratória 360Graus . Isso permite que o acionamento seja girado para a posição operacional ideal e com pouco esforço. - Transmissão de alto desempenho. - Acessórios abrangentes, como: Braços de Reação,  Soquetes de impacto, inserções e suportes estão disponíveis em designs padrão e especiais. - Outros tipos disponíveis mediante solicitação. www.megatorc.com juan@megatorc.com/ comercial@megatorc.com 55+31+3195.6328 / 55+ 31.9.9950.49

  • No alternative text description for this image
Like
Reply

Great edition. Especially exciting to see tooling mature around agents and real-time simulation. This is where developer workflows are heading.

Like
Reply
Gagandeep Singh

🚀 Aspiring AI Engineer | IT Student | Passionate About Machine Learning & AI | Exploring Cybersecurity & Problem-Solving

2d

Thanks for sharing! The part about Salesforce investing in AI training truly highlights how serious the industry is about preparing the workforce for the next wave of innovation.

kushagra sanjay shukla

Masters in Computer Applications/data analytics

2d

Brilliant

Like
Reply

To view or add a comment, sign in

More articles by HackerRank

Explore content categories