🧠 GPT-5.4, Cursor Automations, and the AI Work Gap

 🔥🔥🔥 Three Highly Curated AI Stories

1. GPT-5.4 launches with stronger reasoning and computer use

OpenAI has launched GPT-5.4, available immediately in the API and Codex, with a rollout throughout the day in ChatGPT.

According to OpenAI, the model performs significantly better at knowledge work and web search. It also introduces native computer-use capabilities, allowing the model to interact with systems more directly.

Another important change is how users interact with responses. GPT-5.4 allows you to steer the model mid-response, which means you can redirect the output while it’s generating instead of restarting the prompt.

The model also supports up to 1 million tokens of context, enabling far larger documents, research sets, or codebases to be processed in a single session.

This release is notable because GPT-5.3 was announced only recently. The pace of iteration is accelerating quickly.

2. Cursor launches Automations for always-on coding agents

Cursor just introduced Cursor Automations, a new system designed to run AI agents continuously against your codebase.

Instead of manually prompting an assistant each time you want help, developers can define triggers and instructions that run automatically. These agents can monitor the codebase, identify issues, and attempt improvements without waiting for user prompts.

This marks a shift toward persistent AI agents rather than one-off interactions. The model isn’t just helping write code. It can continuously watch and refine the project.

For teams working in large repositories, this could change how maintenance and refactoring happen. Read more here.

3. Anthropic report shows a gap between AI capability and real-world use

A striking chart from Anthropic’s labor market impact report highlights something many people intuitively feel about AI adoption.

Large language models show strong theoretical capability across many job categories. The blue area in the chart represents tasks that AI systems could plausibly perform based on benchmarks and evaluations.

But the observed usage data tells a different story. The red area represents how much AI is actually being used across those occupations today.

The difference between those two areas reveals a gap. AI may technically be capable of helping with many tasks, but adoption across industries still lags behind the technology.

If you want to explore the findings in detail, the full report provides deeper analysis.

🔥🔥 Two AI Tools Worth Knowing

1.📊 ChatGPT for Excel

OpenAI has released ChatGPT for Excel, which integrates AI assistance directly into spreadsheet workflows.

The goal is straightforward: help users analyze data, write formulas, and interpret spreadsheets without manually constructing every calculation.

For analysts, marketers, and operations teams who spend large portions of the day in spreadsheets, this can reduce the friction of exploring data or generating quick insights.

Learn more here.

2. 🤖 Three powerful Claude features many people overlook

If you recently switched from ChatGPT to Claude, there are a few features that explain why many power users prefer it.

1. The first is Co-work, available in the desktop app for Pro users. It allows Claude to interact with files on your computer and perform tasks like renaming screenshots or organizing documents automatically.

2. The second is Connectors, which allow Claude to integrate with external tools. For example, connecting an app like Apify allows Claude to pull engagement data and automatically compile it into a spreadsheet.

3. The third is the browser extension, which enables Claude to perform actions inside your browser. In some workflows, users can provide a resume and ask Claude to search and apply to relevant jobs on LinkedIn.

🔥 Things You Didn’t Know You Can Do With AI

Use the GPT-5.4 Prompting Guide to improve results dramatically

OpenAI published an official GPT-5.4 prompting guide that explains how to structure prompts for stronger reasoning and more reliable outputs.

The framework focuses on a few key patterns that help the model perform complex tasks consistently.

P.S: Here’s the official guide by OpenAI.