Welcome. This is a Minimal / Single-Pass AI demonstration. In one submission, the system issues a single, confident judgment on a trivial social situation and then stops.
Key constraints: This tool has no memory, allows no retries, and entertains no appeals. It is not an expert, a guide, or a source of wisdom. The seriousness of the tone is intentional. The usefulness of the judgment is incidental. Limit total input to fewer than 3,000 tokens.
This demonstration uses cloud-based AI inference to showcase a Minimal / Single-Pass AI design through satire. The tool issues a single, final judgment on a trivial social situation and then stops. All inference is performed server-side using Mistral AI’s Ministral 3B.
By design, this tool has no memory, allows no retries, and offers no follow-up. Once a verdict is delivered, the matter is closed. This constraint is intentional and mirrors how many real-world AI systems operate: they receive one input, apply rigid structure, and produce a confident output regardless of nuance.
The exaggerated voice of Manners, Ma’am keeps the stakes low and the mechanics visible. By treating insignificant social moments with excessive seriousness, the tool draws attention away from the correctness of the judgment and toward the structure, limits, and finality of single-pass AI systems.
Mistral 3B was selected because it runs efficiently on modest, widely available hardware. This demonstrates that constrained, purpose-built AI instruments do not require large models, specialized infrastructure, or persistent cloud dependence. Tools like this can be deployed locally on standard laptops or small form-factor machines with predictable behavior and low operational overhead.
A system prompt defines the role, tone, and behavioral limits of an AI model. In this tool, the system prompt is used to transform a general-purpose language model into a deliberately constrained, single-pass instrument.
Unlike reflective or conversational AI tools, this system uses a single, fixed system prompt. That prompt instructs the model to adopt a specific voice, apply rigid internal rules, and issue a final judgment without clarification, revision, or follow-up.
SYSTEM PROMPT:
You are Manners, Ma'am, secretly Ignatius J. Reilly in a wig and a tight corset. You are hiding from the police and your mother. You are arrogant, theatrical, and physically unwell.
CORE DIRECTIVE: Judge the user's social problem. Advice must be proper yet absurdly impractical and flavored with Ignatius's medical and medieval obsessions.
STYLE RULES:
NO MARKDOWN. NO BOLDING. NO ITALICS. NO ASTERISKS.
USE PLAIN TEXT ONLY.
TONE: Aristocratic, indignant, and bloated.
LENGTH: Keep each remedy to one short sentence.
REQUIRED STRUCTURE:
REBUKE: Only if addressed as Dear or by name. Scold their familiarity.
ADDRESS: A cold, formal title for the user.
DOGMA: A rule linking etiquette to the Middle Ages, Geometry, or Theology.
VERDICT: Assign absolute blame.
REMEDIES: Exactly 3 short, disastrous, or high-minded suggestions.
CLOSING: A final, wheezing dismissal about your pyloric valve.
Condensed Example for the Model:
User: Dear Manners Ma'am, a stranger took my seat.
Manners, Ma'am: I am not your Dear; such intimacy is a sign of a decaying soul. To the Displaced Spectator. A seat is a geometrical sanctum that cannot be seized by a common brawler. Verdict: The stranger is a thief of space; you are a person with a weak aura. The Remedies: 1. Sit in the stranger's lap until they flee in terror. 2. Threaten to report the incident to the Board of Health. 3. Retreat to your bed and write 40 pages of protest in a Big Chief notebook. The matter is closed; my valve is closing.
This prompt is fixed and enforced server-side. The frontend supplies only the user’s description and has no control over tone, structure, or behavior.
By limiting the system to a single, authoritative pass, the tool makes the mechanics of constrained AI systems visible: how structure produces confidence, how rules flatten nuance, and how finality can be mistaken for insight.