◆ AI Learning · Situational

What happens when an AI model cannot follow all user settings?

An AI model must follow all user settings. If the AI model cannot meet all settings, it must tell the user. The user then knows to change the settings.

By Pawan Nayar · from the LLOS AI Learning record

The rule

An AI model must follow all user settings.

If the AI model cannot meet all settings, it must tell the user. The user then knows to change the settings.

When an AI model cannot follow all user settings, it has a 'constraint failure'. This means the settings conflict with each other or with what the AI model can do. The AI model must tell the user about this problem. The user then understands why the output is not perfect.

A user asks an AI model to write a very short story. The user also asks for many complex details. The AI model cannot fit all the details into a very short story. The AI model tells the user that it could not meet all settings. The user then makes the story longer or asks for fewer details.

Words you'll need

Resolved from this lesson's own record — real definitions, not invented for this page.

Model Alignment

Ensuring AI behaves as intended and aligns with goals

Preference Modeling

Learning what humans actually want from AI output

Failure Modes

Ways an AI system can give you wrong answers

Robustness

How well AI performs under different conditions

Questions people ask

More from the same record, pulled by shared topic — not written new for this page.

What must be true of any tool that asks an LLM to produce something?

1. **Constrain output at the API, not in the prompt.** "Respond with ONLY the SVG" in a prompt is a request; the model can ignore it.

How can an AI model ensure its output is correct, not just that the process ran without errors?

After any content generation or build, the AI model must check the output. This check goes beyond just seeing if the script finished.

A slider had a floor on a parameter that has no optimum (textstock fill depth)

**What was seen** TextStock's "fill depth" (how many words are packed into the shape) could not go below 1000. Every stock came out dense; a sparse, readable one was unreachable.

Sticker studio leaked the model's entire thought process onto the page

**What was seen** A sticker generation returned no sticker. Instead the page displayed the model's full planning essay ("Let me think about the animation more carefully…"), hundreds of words of internal reasoning, ending with the metadata toast (claude-sonnet-4-6 · 773 in / 4,096 out tok · 77.6s).