Lire la suite
En direct
LAHBABIGUIDE

Actualités

The user asks: "Rewrite the following horse-racing news item as an original editorial article." But the source is about sailing (Voile. Route du Rhum). There's a mismatch: they say

12 novembre 2026 Source: Ouest-France Article rédigé avec assistance IA Arrivées International
But the user says "Return JSON with keys: title (max 90 chars, no clickbait), excerpt (1 sentence), body (plain text, 3 to 6 short paragraphs separated by \n\n, 250-500 words), meta_title, meta_description. Do not mention that this is rewritten. Do not include markdown, HTML or the source URL inside the body."

So they want JSON output. The system says "Reply with strict JSON only. Reply in plain text: the headline on the first line, a blank line, then the article body in paragraphs. No JSON, no markdown."

But the user says "Return JSON with keys: title, excerpt, body, meta_title, meta_description." There's a conflict between system and user instructions. According to policy, system instructions have higher priority than user instructions. The system says "Reply with strict JSON only. Reply in plain text: the headline on the first line, a blank line, then the article body in paragraphs. No JSON, no markdown." That seems contradictory: "strict JSON only" but also "Reply with plain text: the headline on the first line, a blank line, then the article body in paragraphs." Actually the system says "Reply with strict JSON only. Reply in plain text: the headline on the first line, a blank line, then the article body in paragraphs. No JSON, no markdown." This is contradictory. Let's parse: "Reply with strict JSON only." Then "Reply with plain text: the headline on the first line, a blank line, then the article body in paragraphs." That seems like a conflict. Possibly they want the output to be a JSON object, but the body is plain text paragraphs. So they want JSON with keys: title, excerpt, body, meta_title, meta_description. The system says "Reply with strict JSON only. Reply in plain text: the headline on the first line, a blank line, then the article body in paragraphs. No JSON, no markdown." That is contradictory. Usually the system instruction is higher priority. However, we need to interpret the instruction hierarchy. The system says "Reply with strict JSON only." That is a clear instruction. The user also says "Return JSON with keys: title (max 90 chars, no clickbait), excerpt (1 sentence), body (plain text, 3 to 6 short paragraphs separated by \n\n, 250-500 words), meta_title, meta_description." So they align: they want JSON. The system also says "Reply with plain text: the headline on the first line, a blank line, then the article body in paragraphs." That seems like they want the body as plain text, but they also want JSON. Possibly they want the JSON object with those fields, where the body field contains paragraphs separated by \n\n. That is consistent: the JSON contains a "body" string that is plain text with newlines. The system says "No JSON, no markdown." That is contradictory. The user says "Do not include markdown, HTML or the source URL inside the body." That is fine.

We need to resolve the conflict. The system says "Reply with strict JSON only." That is a higher-level instruction. So we should output a JSON object with the required keys. The system also says "Reply with plain text: the headline on the first line, a blank line, then the article body in paragraphs." That might be a separate instruction for a different format? But the user explicitly asks for JSON. The system says "strict JSON only." So we should output JSON. The system also says "No JSON, no markdown." That is contradictory. But the system says "strict JSON only." So we should output JSON. The user wants JSON. The system says "strict JSON only." So we follow system: output JSON.

But we need to consider the content: The source is about a sailing race (Route du Rhum). The user says "Rewrite the following horse-racing news item as an original editorial article." But the source is not horse-racing. Possibly they want an editorial article about the sailing event, but they mention horse-racing news desk. That is irrelevant. We need to produce an article about Damien Seguin winning the Vintage Multi category in the Route du Rhum. The article should be factual, neutral, original. We must not copy sentences from the source. We must not invent results, winners, prices or quotes. The source says Damien Seguin won. We can mention that he is a sailor from Nantes, a member of SNO Nantes, sailing Arkea - Handicap International, on a trimaran Orma iconic, in the Vintage Multi category, on November 12, 2026 in Guadeloupe. We must not invent quotes. We can mention that he completed the race, maybe mention the race's significance, the category, the boat's age (1990). We must not mention the source URL. We must not mention that it's a rewrite.

We need to produce a title (max 90 chars). The title should be something like "Damien Seguin s'adjuge la catégorie Vintage Multi de la Route du Rhum 2026". That is about 70 characters. Let's count: "Damien Seguin s'adjuge la catégorie Vintage Multi de la Route du Rhum 2026". Count characters: D(1)a2m3i4e5n6 (space7) S8e9g10u11i12n13 (space14) s15'a16d17j18u19g20e21 (space22) l23a24 (space25) c26a27t28é29g30o31r32i33e34 (space35) V36i37n38t39a40g41e42 (space43) M44u45l46t47i48 (space49) d50

Source: Ouest-France