So really I think "follows prioritised instructions" might be a better framing, where the system prompt is taken as the highest priority instruction?
PROMPT ENGINEERING
-
Redefining AI Misuse Requests: Safety vs Developer Intent
By
–
"requests for misuse" only makes sense, to me, if defined as "request to do a thing that the system prompt asks not to do". This is not necessarily harmful or unsafe. It's just not what the developer wanted.
-
User Intent Framing and AI Safety Safeguards Context
By
–
I think I disagree with your framing. "The user is intentionally asking for something that could lead to harmful outcomes if no safeguards are in place" could refer to no prompts, or all prompts, depending on context.
-
AI Agent Architect Newsletter: Bridging Technical and Real-World Impact
By
–
The AI Agent Architect newsletter helps professionals bridge the gap between the technical aspects of AI Agents and real-world impact, without needing to code. Sign up here: https://
theagentarchitect.substack.com -
Seedance 1 Pro Model Behavior Reproducibility with Image Prompts
By
–
I wish I could reproduce this reliably with a prompt. Right now it seems to be a behaviour found with certain start images and "aggressively mediocre home footage" as a prompt.
— fofr (@fofrAI) 15 juillet 2025
This is with Seedance 1 Pro via my `not-real` model. https://t.co/qE1hE9SWiL pic.twitter.com/lXXIwWgSXQI wish I could reproduce this reliably with a prompt. Right now it seems to be a behaviour found with certain start images and "aggressively mediocre home footage" as a prompt. This is with Seedance 1 Pro via my `not-real` model.
-
System Prompt Changes Cannot Fix Underlying Model Bias
By
–
This is the key issue – changes to the system prompt can only paper over the underlying behavior that appears to have been trained into the model. @simonw those system prompt changes you highlighted won't really fix the underlying bias.
-
LLM Simulated Environments for Agent Training Without Real Tools
By
–
By this, I mean something like, instead of creating an environment for the model to use during training, prompt a LLM to output what it thinks the environment would output for any given tool call.
-
Custom System Prompts via API Implementation
By
–
Presumably yes – using the API you get to provide your own system prompt
-
Detecting LLM Hallucinations: Consistency as a Key Indicator
By
–
I find those are usually easy to detect, because you don't reliably get the same hallucinated guidelines back multiple times
-
Prompt Injection Attacks: Security Concerns and Data Protection Measures
By
–
I'm too scared of prompt injection attacks to give it that level of access! Do you have anything in place to make sure nobody sneaks malicious instructions into your setup that steal your data?