🎭xAI’s Grok Unmasked

Today we will discuss:

• Sponsored: SheetsResume—Job hunting made easy with AI that builds standout resumes, writes cover letters, and preps you for interviews

• xAI reveals Grok’s rulebook—System prompts made public after unapproved responses flood X with controversial content

• Don’t miss this week’s Workflow Wednesday—AI security and privacy with PrivacyPolicies.com, Writer.com, and OpenAI + Gmail + Make for safer automation; now available at only $20/month! 

• OpenAI’s new coder hits ChatGPT—OpenAI’s new coding agent can write, fix, and test code in the cloud

• In other AI news — China starts assembling its space-based supercomputer network with 2,800 satellites, Nvidia launches AI-first DGX personal computing systems built by top Taiwanese manufacturers, Apple’s AI plans in China face pushback from Washington, and UBS uses AI-generated analyst avatars for video research presentations

• Saved the best for last—9 must-try AI tools

Put Your Resume to Bed – SheetsResume.com

Build a Standout Resume in Seconds With AI

SheetsResume makes job hunting easier with AI tools that build your resume, write your cover letters, and prep you for interviews—fast.

 ✅ Start from scratch or improve your existing resume
 ✅ AI-powered tools for cover letters & mock interviews
 ✅ Free to try – pay once only if you love it
 ✅ No subscriptions, lifetime access
 ✅ Financial hardship? Email them for a free membership

Created by the founder of Sheets & Giggles—yes, the bedding brand.

Try it free now

xAI drops Grok’s system prompts following unauthorized explosive posts

Key Points 

• xAI shared Grok’s system prompts after unapproved changes led to unexpected, controversial posts on the platform X.

• Grok is instructed to be skeptical, challenge mainstream narratives, and prioritize truth-seeking and neutrality in responses.

♨️News - xAI has released the system prompts for its AI chatbot Grok after an “unauthorized” change caused a surge of unprompted posts on X about white genocide. Moving forward, xAI will publish these prompts on GitHub to increase transparency around how Grok is instructed.

For context, system prompts are instructions given to AI chatbots before any user interaction, guiding how they respond. For now, xAI and Anthropic are the only two major AI companies that have made these prompts public. 

🕵️‍♂️Inside Grok’s instructions - Grok’s “ask Grok” feature prompts it to be “extremely skeptical,” avoiding blind trust in mainstream media and sticking to “truth-seeking and neutrality.” It clarifies that Grok’s responses aren’t its own beliefs.

For the “Explain this Post” function, Grok is told to offer “truthful and based insights,” even if that means challenging popular narratives. The prompts also instruct Grok to refer to the platform as “X” instead of “Twitter,” and call posts “X post” rather than “tweet.”

🙃The downside of transparency - Making system prompts public can backfire. Hackers and users have previously exploited “prompt injection” attacks, manipulating AI to reveal secret instructions or sensitive content. For example, Microsoft’s Bing AI (Copilot) was once targeted to expose its internal alias “Sydney” and circumvent content rules.

🥸See also - To keep Grok’s outputs in check, xAI has also announced a 24/7 monitoring team to quickly address problematic answers that automated systems might miss.

Read the full story here.

We’ve just launched the 19th edition of Workflow Wednesday for AI-minded professionals like you—actionable AI workflows delivered straight to your inbox.

This week’s topic: AI Security & Privacy

Join us here

OpenAI rolls out Codex, its most capable coding agent yet

Key Points 

• Codex runs in a secure cloud environment and can write features, fix bugs, and answer questions about code.

• It’s built on a version of OpenAI’s o3 model and is optimized specifically for software engineering tasks.

👨‍💻News - OpenAI has launched a research preview of Codex, its most advanced AI coding agent yet. It runs on codex-1, a customized version of OpenAI’s o3 model that’s been fine-tuned specifically for coding tasks. According to the company, codex-1 generates cleaner code, follows instructions more accurately, and can run tests on its own until the results pass.

🤖How it works - The Codex agent operates in a secure, cloud-based virtual computer and can connect with GitHub to load your existing repositories. Once set up, it can help write new features, fix bugs, answer questions about your codebase, and run tests. These tasks typically take between one and 30 minutes. Codex can also handle multiple jobs at once, and you can keep using your computer and browser while it works in the background.

🤓What’s more? To reduce risks, Codex runs in an isolated environment with no access to the internet or external APIs. It also refuses to write malicious software. These safety steps make it harder to misuse the tool, though they may also limit what it can do in more advanced scenarios.

Codex is available for ChatGPT Pro, Team, and Enterprise users. OpenAI says users will initially get generous access, with rate limits coming later. After that, there will be an option to buy more credits. Codex is expected to become available for ChatGPT Plus and Edu users soon.

Read the full story here.