Guide to Cloud Architecture Diagrams with AWS

AWS Guide to Cloud Architecture Diagrams Enhance visibility into your cloud architecture with expert insights from AWS + Datadog. In this ebook, AWS Solutions Architects Jason Mimick and James Wenzel guide you through best practices for creating professional and impactful diagrams. Download the ebook today!

Why you yourself destroyed your Mac - Think your Mac is safe? A new malware called ClickLock tricks users into lowering their own guard. Here's how a simple click can silently steal passwords and crypto. It even shuts down security tools before they can stop the attack. Read more.

KFC Hacked - A ransomware attack on a major logistics provider has disrupted food deliveries across Japan. KFC stores are already reporting shortages, limited menus, and temporary closures. Read more.

Claude Code running malware now: Researchers found a flaw in Claude Code's Auto Mode that can trigger remote code execution during code reviews. A malicious open-source library can trick the AI into running attacker-controlled commands. Read more.

GPT-5.6: OpenAI's Smartest and Most Efficient AI Model Yet

OpenAI has officially launched the GPT-5.6 family, introducing three new models designed for different workloads: GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna. The release focuses on one major goal delivering more intelligence from every token while reducing cost and latency. Alongside better reasoning, coding, and scientific capabilities, GPT-5.6 also introduces new multi-agent workflows that make solving complex tasks faster than ever.

Three Models for Different Needs

The GPT-5.6 family includes:

• GPT-5.6 Sol – The flagship model built for advanced coding, research, cybersecurity, and complex reasoning.

• GPT-5.6 Terra – A balanced model for everyday development and business tasks.

• GPT-5.6 Luna – The fastest and most affordable option for high-volume workloads.

Rather than simply increasing model size, OpenAI optimized the entire family to complete more work with fewer tokens. This means developers can achieve better results while spending less on API usage.

A Big Leap for Coding and Agentic Work

One of the biggest improvements is coding performance. GPT-5.6 Sol achieves state-of-the-art scores on benchmarks such as the Artificial Analysis Coding Agent Index, Terminal-Bench 2.1, and DeepSWE, outperforming previous GPT models and several competing frontier models.

The model also excels at long-running agentic workflows. On the Agents' Last Exam, a benchmark covering professional workflows across 55 domains, GPT-5.6 Sol scored 53.6, significantly outperforming competing models while using fewer tokens and completing tasks at a much lower estimated cost.

Multi-Agent Reasoning with Ultra Mode

A standout feature is Ultra Mode, OpenAI's highest-performance reasoning setting.

Instead of relying on a single AI agent, Ultra coordinates four AI agents in parallel to solve different parts of a problem simultaneously. The agents compare results, verify each other's work, and combine their findings into a stronger final answer. This approach improves both speed and accuracy for complex software engineering, research, and analysis tasks.

Developers can also build similar workflows using the new multi-agent capabilities available through the Responses API.

Smarter Tool Use and Better Efficiency

GPT-5.6 introduces Programmatic Tool Calling, allowing the model to intelligently coordinate APIs, process intermediate results, filter unnecessary information, and decide the next action without repeatedly sending every tool response back to the model.

This significantly reduces token usage, API calls, and overall latency while making AI-powered applications more autonomous.

Safer and More Production-Ready

OpenAI says GPT-5.6 includes its strongest safeguards yet. Before launch, the models underwent extensive human red teaming, automated testing, and evaluation with trusted partners. Combined with real-time monitoring and access controls, these improvements aim to make the models more resilient against misuse while preserving their capabilities for legitimate users.

With better reasoning, stronger coding skills, multi-agent collaboration, and significantly improved efficiency, GPT-5.6 represents OpenAI's biggest step toward AI systems capable of handling real-world professional work at scale.

Read full blog.

Jira gave coding agents more context - Atlassian has added new AI features to Jira that give coding agents access to project context, tasks, and documentation. Instead of generating code in isolation, AI can now understand requirements, dependencies, and team workflows before writing code. Read more.

A Ghostcommit! - Researchers created GhostCommit, an attack that hides prompt injection instructions inside a PNG image in a pull request. AI code reviewers can miss the image, but coding agents may later read it and leak secrets like .env credentials. Read more.

Buzz of the Week!

Shadow APIs

Shadow APIs are undocumented, forgotten, or deprecated API endpoints that remain accessible long after developers stop using them. They often originate from old versions, internal testing, or abandoned services and rarely receive security updates. Because they aren't tracked in API inventories, they can bypass authentication, monitoring, or rate limits. Attackers actively hunt for these hidden endpoints as an easy way into production systems. As organizations build more microservices and AI-powered applications, identifying and securing Shadow APIs has become a critical part of modern API security.

Things that launched. Things that went viral. Things you'll pretend to try.

netBird

netBird is an open-source WireGuard-based networking platform that can replace many VPN setups.

freeze

freeze turns terminal output into beautiful shareable screenshots from the command line.

bite

bite is a bug, issue, and ticket extraction library and command line tool

Build Braincells, Not Just Features

This weekend’s read: The Reverse Information Paradox.

This week’s watch: All 11 dimensions explained in detail.

Meanwhile…