/AI News/Issue 159

Issue #159 14 stories

Global AI Weekly

A sneak peek of OpenAI's first gadget

Published Tuesday, July 21, 2026

In this issue

Highlights

3 stories
Microsoft's Nadella criticizes Anthropic's Fable for being "editorially controlled"
cnbc.com

Microsoft's Nadella criticizes Anthropic's Fable for being "editorially controlled"

Microsoft CEO Satya Nadella criticized Anthropic’s Claude Fable safeguards, arguing that unpredictable refusals and model-switching make the tool overly “editorially controlled.” Speaking to Copilot engineers, he said enterprises should be able to customize AI with proprietary data while retaining control, rather than accepting restrictions imposed by foundation-model providers.

In this issue

Research

2 stories
Pretraining Data Can Be Poisoned through Computational Propaganda
arxiv.org

Pretraining Data Can Be Poisoned through Computational Propaganda

Researchers show attackers may poison future language models by mass-posting malicious text through ordinary web comments. Their HalfLife analysis tracks whether injected content survives crawling, extraction and filtering into training corpora. The work suggests public discussion systems could become an overlooked supply-chain attack on AI models before training even begins.

In this issue

Video

2 stories
Coding smarter with GitHub Copilot
youtube.com

Coding smarter with GitHub Copilot

Are you still wondering what the point is of custom instructions and skill files? Or how to use them in the best way? Let's find out through live demos. In this session, we’ll see how GitHub Copilot can support our workflow in Visual studio Code. You may have seen the highlights before, so we’ll focus on what matters: how to get consistently better results, instead of feeling like you’re explaining your code to a toddler. Because when you give GitHub Copilot the right instructions, it will become a tool that actually saves time, comes up with options you might miss, and occasionally surprises you in a good way.

Codex just got better for developers
youtube.com

Codex just got better for developers

OpenAI demonstrates how Codex now handles more of the software-development lifecycle: investigating issues, planning changes, writing and testing code, and preparing review-ready pull requests. The update emphasizes longer-running, parallel agent workflows, deeper project context, and stronger developer control, helping teams tackle routine fixes, complex refactors, and migrations faster.

In this issue

Articles

3 stories
How does Chain-of-Thought (CoT) Prompting work?
outcomeschool.com

How does Chain-of-Thought (CoT) Prompting work?

In this blog, we will learn about how Chain-of-Thought (CoT) Prompting works. We will also see why a model that jumps straight to the answer often gets it wrong, how making it reason step by step fixes this, the difference between zero-shot and few-shot CoT, and where this technique is truly useful.

In this issue

Upcoming Events

1 story
Upcoming Events Around the World
globalai.community

Upcoming Events Around the World

From local meetups to full-day conferences, the Global AI Community brings developers, architects, founders, and AI enthusiasts together to learn, build, and connect. Discover hands-on workshops, expert talks, community gatherings, and flagship events like AgentCon, AgentCamp, and Global AI Construct. Whether you are exploring AI for the first time or building production-ready agentic systems, you will find opportunities to sharpen your skills, meet like-minded people, and become part of a thriving global community.

In this issue

Code

2 stories
Codex Bootcamp
academy.openai.com

Codex Bootcamp

Codex Bootcamp is a live OpenAI Academy series for technical builders and engineering teams looking to use Codex more effectively across real development workflows. Explore the 101, 201, and 301 sessions and register for the levels most relevant to you.

In this issue

Podcast

1 story
Does Claude Have Private Thoughts?
open.spotify.com

Does Claude Have Private Thoughts?

Cal Newport examines Anthropic’s research into a hidden “global workspace” inside Claude, asking whether internal representations amount to private thought. He explains the mechanics behind the findings, separates intriguing evidence from consciousness hype, and argues that the work reveals something important about model computation—without proving that Claude is secretly thinking.

Every Tuesday

Get the next issue in your inbox.

The AI links worth your time, curated by the community and always free.