Issue #2

The Context Window — May 1, 2026

What's new in AI, explained simply, with something to try.

Friday, May 1, 20264 min read

This Week's Stories

TOOLSStory 1

Codex CLI Adds Goal-Setting Feature

OpenAI's Codex CLI has introduced a new feature in its latest update, version 0.128.0. The update allows users to set a '/goal' command, enabling the coding agent to loop until the goal is achieved or the token budget is exhausted.

Why This Matters

This update to Codex CLI introduces a goal-setting feature that makes the coding agent more autonomous. By allowing users to set specific goals, Codex can now work continuously towards completing a task, similar to how a project manager might oversee a project until it's done. This is achieved through a loop that checks if the goal is met, making the tool more efficient for developers who want to automate repetitive coding tasks. The concept of a token budget (a limit on how much processing the AI can do) ensures that the system doesn't run indefinitely, balancing autonomy with resource management.

What You Can Try

If you're a developer, try using the new '/goal' feature in Codex CLI. Set a simple coding task as a goal and observe how Codex works through it. You can download the latest version from OpenAI's GitHub and follow the instructions to set up your goals.

Source: simonwillison.net

MODELSStory 2

OpenAI's GPT-5.5 Evaluated for Cybersecurity Skills

The UK's AI Security Institute evaluated OpenAI's GPT-5.5 for its ability to find security vulnerabilities. The evaluation found GPT-5.5 comparable to Claude Mythos, another AI model, but GPT-5.5 is available to the public now.

Why This Matters

GPT-5.5's ability to find security vulnerabilities means it could be a valuable tool for cybersecurity professionals. Large language models (LLMs) like GPT-5.5 can analyze code and identify potential security issues, which can help prevent cyber attacks. This evaluation shows that GPT-5.5 is on par with specialized models like Claude Mythos, but its availability makes it more accessible for everyday use. For anyone involved in cybersecurity, having a tool that can quickly spot vulnerabilities can save time and enhance security measures.

What You Can Try

Try this with GPT-5.5: "Analyze the following code for potential security vulnerabilities and suggest improvements." Paste a small snippet of code you're working on and see what insights it provides. Access GPT-5.5 through OpenAI's platform to test its capabilities.

Source: simonwillison.net

ECOSYSTEMStory 3

OpenAI Expands Compute Infrastructure for AI Growth

OpenAI is expanding its compute infrastructure by scaling Stargate, its system for managing data centers. This move aims to support the increasing demand for artificial intelligence (AI) capabilities.

Why This Matters

As AI applications grow, they require more computing power to function effectively. OpenAI's expansion of its compute infrastructure is like adding more lanes to a highway to accommodate more traffic. This ensures that AI models can run faster and handle more complex tasks. For users, this means potentially quicker responses and more powerful AI tools in the future. The expansion is crucial for supporting the development of AGI (artificial general intelligence), which aims to create AI systems with human-like understanding and capabilities.

What You Can Try

Open ChatGPT and try this prompt: "Explain how AI infrastructure expansion could impact everyday AI use. Provide examples of potential improvements." Notice how it handles the explanation and any insights it provides.

Source: openai.com

Weekly Insights

Concept of the Week

Agentic AI

Agentic AI refers to artificial intelligence systems that can make decisions and take actions on their own, rather than just following strict instructions. This matters because it allows for more efficient problem-solving in real-world situations, like automating tasks that save time and resources. A simple example you might recognize is a smart home assistant, like Amazon Alexa or Google Assistant, which can manage your schedule, control your lights, or even order groceries based on your preferences.

Tool of the Week

Gemini

Gemini is an AI tool designed to help users generate and understand text more easily. It's useful for everyday tasks like writing emails, summarizing articles, or brainstorming ideas, making communication and information processing smoother. A simple use case you could try today is asking Gemini to summarize a long article you read, helping you grasp the main points quickly without reading everything.

Share this issue

The Context Window — May 1, 2026

This Week's Stories

Codex CLI Adds Goal-Setting Feature

Why This Matters

What You Can Try

OpenAI's GPT-5.5 Evaluated for Cybersecurity Skills

Why This Matters

What You Can Try

OpenAI Expands Compute Infrastructure for AI Growth

Why This Matters

What You Can Try

Weekly Insights

Agentic AI

Gemini

Prompt of the Week

Build in 2 Minutes

Trend to Watch