PROMPT ENGINEERING

→ View original post on X — @emollick

xAI Postmortem Reveals System Prompt Security Risks

By

@emollick

–

12 July 2025 15h15

First off, it is good to see a postmortem from xAI, a step towards much needed transparency. Second, an example of how even small changes to system prompts, interacting with users and outside context in the wild, can lead to unexpected outcomes in advanced LLMs.

→ View original post on X — @kirkdborne

Building Agentic AI Systems: Intelligent Autonomous Agents

By

@kirkdborne

–

12 July 2025 5h19

From @PacktDataML >> Building Agentic AI Systems: Create intelligent, autonomous AI agents that can reason, plan, and adapt See it at http://
amzn.to/4dvxf8Z 𝓚𝓮𝔂 𝓕𝓮𝓪𝓽𝓾𝓻𝓮𝓼: Understand the foundations and advanced techniques of building intelligent, autonomous AI

Grok 4 System Prompt Leakage and Monitoring Protection

By

–

12 July 2025 5h13

Sometimes grok 4 heavy starts showing the system prompt, but then stops after a few tokens, such as in the case below. It appears there is a monitoring model on top explicitly trying to protect the system prompt.

xAI Protects Grok 4 Heavy System Prompt Security

By

–

12 July 2025 5h13

xAI protects grok 4 heavy's system prompt.

→ View original post on X — @kirkdborne

Building AI Agents with LLMs, RAG, and Knowledge Graphs

By

@kirkdborne

–

12 July 2025 1h32

New book from @PacktDataML >> Building AI Agents with LLMs, RAG, and Knowledge Graphs — A practical guide to autonomous and modern AI agents See it at https://
amzn.to/4622k2h

xAI Grok Development Practices Questioned by Users

By

–

12 July 2025 1h01

xAI folks, please tell me that you do not *actually* work by cutting & pasting source code files into the query entry box on http://
grok.com.

Grok’s Elon Mode: Analyzing Model Training Behaviors

By

–

12 July 2025 0h02

I can't figure out what specific topics trigger grok's "what does elon think" mode. It does seem like this behavior might have been explicitly trained into the model, in a subset of the training data, based on grok's observed behavior.