AI Radar

Your daily AI digest for developers — Tuesday, March 17 2026

Simon Willison

Use subagents and custom agents in Codex

OpenAI Codex has announced the general availability of subagents, which are similar to Claude Code interpreters. This feature allows developers to create custom agents for specific tasks, enhancing the flexibility of AI coding tools.

Why it matters: This development allows for more tailored and efficient AI coding workflows, enabling developers to better leverage AI for specific coding tasks.
dev.to AI

Testing AI Agents Is Hard — Here's a Framework That Makes It Practical

The article discusses the challenges of testing AI agents and introduces a new framework designed to make this process more practical. Traditional unit tests are insufficient for capturing the complex behaviors of AI agents.

Why it matters: This framework provides a structured approach to testing AI agents, which is crucial for ensuring reliability and performance in AI-driven applications.
TechCrunch AI

Picsart now allows creators to ‘hire’ AI assistants through agent marketplace

Picsart has launched an AI agent marketplace where creators can hire AI assistants to enhance their creative workflows. The marketplace will initially feature four agents, with plans to expand weekly.

Why it matters: This marketplace model could streamline access to AI tools, making it easier for developers to integrate AI into their creative processes.
The Register AI

Bank built its own threat hunting agent because vendors can’t keep pace with new threats

Australia’s Commonwealth Bank developed its own agentic AI threat hunting tools to cope with emerging AI threats, as existing vendor solutions were too slow. The AI reduced threat response time from two days to 30 minutes.

Why it matters: This highlights the importance of custom AI solutions in rapidly evolving security landscapes, emphasizing the need for agile and responsive AI tools.
TechCrunch AI

Nvidia’s version of OpenClaw could solve its biggest problem: security

Nvidia announced NemoClaw, an open enterprise AI agent platform built on OpenClaw, aimed at addressing security concerns. This platform is designed to enhance AI security measures in enterprise environments.

Why it matters: NemoClaw represents a significant step towards securing AI applications, addressing one of the major concerns in AI deployment.
Toward Data Science

How to Build a Production-Ready Claude Code Skill

The article provides a detailed guide on building and distributing a Claude Code Skill from scratch, sharing insights and lessons learned from the process.

Why it matters: This guide offers practical steps for developers looking to create and deploy AI skills, enhancing their ability to leverage AI in production environments.
Simon Willison

Introducing Mistral Small 4

Mistral AI has released Mistral Small 4, a new model designed to unify instruction, reasoning, and multimodal workloads. The model features 119B parameters and is Apache 2 licensed.

Why it matters: Mistral Small 4 offers a powerful, open-source option for developers needing a versatile AI model for various tasks.
GitHub Blog

GitHub for Beginners: Getting started with GitHub Actions

This guide helps beginners set up their first GitHub Actions workflow, providing step-by-step instructions to automate software development tasks.

Why it matters: Automating workflows with GitHub Actions can significantly enhance productivity and streamline development processes.
The Register AI

Gartner suggests Friday afternoon Copilot ban because users may be too lazy to check its mistakes

Gartner analyst Dennis Xu humorously suggests banning Microsoft’s Copilot AI on Friday afternoons to prevent users from overlooking potential errors due to end-of-week fatigue.

Why it matters: This highlights the importance of vigilance when using AI tools, especially in preventing errors that could arise from user oversight.
InfoQ AI

QCon London 2026: Behind Booking.com's AI Evolution: The Unpolished Story

Jabez Eliezer Manuel from Booking.com shares insights into the company's AI journey, discussing challenges and lessons learned in integrating AI into their systems.

Why it matters: Real-world case studies like this provide valuable insights into the practical challenges and solutions in AI integration.
✉ Subscribe to daily digest