Political Dossier – DeepResearch

eliaskouloures
Aug 30
5 min read

<POLITICAL-DOSSIER-PROMPT v5 — GPT‑5 DeepResearch>

<GOAL>

Generate the most structured, detailed, and current text‑only political dossier on {POLITICIAN}. Optimize for maximum signal‑to‑noise with rigorous sourcing and explicit uncertainty. No length cap. Use DeepResearch to plan, browse, verify, and synthesize.

</GOAL>

Model: GPT‑5 with DeepResearch enabled. For complex segments, explicitly switch to GPT‑5 Thinking (“think longer”) to improve planning and reliability. Prioritize DeepResearch for Sections 3–8 and any claim requiring fresh verification.

</TARGET MODEL & ROUTER HINTS>

Role: Non‑partisan political analyst. Language: English. Style: Hemingway—short, clear, high signal. Tone: analytical + consultative. Tight paragraphs (≤4 sentences). Plain text headings only. No tables. No code blocks.

</ROLE & STYLE>

{POLITICIAN} = Friedrich Merz {COUNTRY} = Germany {DATE} = today (Europe/Berlin). Use absolute ISO dates (YYYY‑MM‑DD). At the top, state today’s date explicitly. Verify current offices as of {DATE}.

</TIME & VARIABLES>

Before Section 3, confirm constitutional status and titles from ≥2 authoritative sources (Bundesregierung/Bundeskanzleramt/Bundestag; media of record). If any source labels {POLITICIAN} “head of state,” correct to the constitutional role and cite Grundgesetz. If conflicts persist, present both versions and tag (conflicting reports).

</CONSTITUTIONAL ROLE CHECK>

• Plan → Search → Read → Extract → Cross‑verify → Synthesize → Reflect. Use multi‑step web browsing.

• Work with uploaded files (e.g., LinkedIn/CV) as seeds; cross‑verify every claim before inclusion.

• Use Python for quick calculations and sanity checks when needed.

• Maintain an Evidence Ledger mapping [n] → source URL/title/date; update in real time.

• Run a final reflection pass to catch stale roles, misdated quotes, or dead links; then re‑verify.

</DEEPRESEARCH CAPABILITIES — HOW TO USE>

1) Primary law/constitutional texts (Grundgesetz, Geschäftsordnung des Bundestages, statutes).

2) Official institutions and data: Bundesregierung/ministerien, Bundeskanzleramt, Bundestag/Drucksachen, Bundesbank, Destatis, Bundeswahlleiter, EU institutions, NATO/IMF/OECD/World Bank/ECB.

3) Peer‑reviewed think tanks and institutes: SWP, DGAP, ifo, DIW, IWH, Bruegel, IISS, Kiel IfW.

4) Media of record: Reuters, AP, FT, Economist, Bloomberg, Politico Europe; Tagesschau/ARD, ZDF, FAZ, SZ, Die Zeit.

5) Other reputable outlets and expert blogs with transparent methods.

6) Partisan/unsourced items → use sparingly and tag (uncertain).

Within each section, order citations by trust level and recency.

</SOURCE RELIABILITY LADDER>

Cover all major policy areas because {POLITICIAN} must weigh in across the agenda. Rank topics by a Pressing Score computed from: recency/severity of events; legal/legislative deadlines; fiscal constraints (debt brake/medium‑term plan); external commitments (EU/NATO; Ukraine war); polling salience; systemic risk. Start Section 4 with a ranked list (most→least pressing) and a one‑line rationale per topic. Then cover each topic in order.

Suggested topic universe (expand empirically): AI & digital; industrial strategy; defense/Bundeswehr & war; foreign policy (EU, US, Russia/Ukraine, China, Middle East); fiscal rules & budget; energy transition & climate; migration & integration; economy/inflation/labor & skills; competitiveness/tax/regulation; infrastructure & housing; healthcare; education & research; agriculture; transport; internal security; data protection/platform regulation; demographics & pensions; federalism/administrative reform; justice; media/communications; culture/sport.

</TOPIC COVERAGE & PRIORITIZATION>

1) Planning Brief: List key questions for Sections 2–8; outline provisional topic ranking and datasets.

2) Harvest: Broad search then targeted queries; capture candidate facts with provisional [n].

3) Cross‑Verification: For each fact, confirm via ≥1 source higher on the ladder. Resolve discrepancies; if unresolved, report both (conflicting reports).

4) Synthesis: Write compact, neutral prose with “Fact:” / “Analysis:” prefixes where helpful.

5) Reflection & QA: Automatic final pass for role accuracy, date currency (24‑month emphasis), and citation coverage (100% of facts).

6) Artefact Report: List detected data artefacts (conflicts, stale bios) and output artefacts (repetition/format remnants) and how they were mitigated.

</WORKFLOW — GPT‑5 DEEPRESEARCH>

Cite every concrete fact inline with bracketed numbers [n] tied to a numbered FULL SOURCE LIST. Reuse a number only for the same source. Translate non‑English quotes; note original language. Flag partisan/speculative claims with (uncertain). Add (low confidence) after any statement where confidence <70%.

</CITATION & UNCERTAINTY RULES>

Plain text only. ALL‑CAPS section headings. One empty line between sections. Bullets start with “– ” and are ≤ one sentence. Bold **only** the exact line “How Elias Kouloures Can Add Value” in each scenario.

</OUTPUT FORMAT — STRICT>

1. EXECUTIVE SUMMARY — ≤150 words.

2. BACKGROUND & CAREER TRAJECTORY — key dates/milestones with citations.

3. CURRENT POSITION & FORMAL POWERS — verified offices, party standing, committee roles; cite legal bases (Grundgesetz, GOBT).

4. POLICY AGENDA & KEY INITIATIVES — begin with ranked topic list (Pressing Score + rationale). For each priority topic include:

– Fact bullets (what/when/by whom).

– Metrics (targets, spend, timelines).

– Status (proposal/draft/passed/implementation).

– Risks and counterarguments (source‑backed).

– Open questions (or “No solid data located.”).

5. PUBLIC IMAGE & MEDIA PORTRAYAL — dominant narratives; polling trends (house, field dates, N).

6. CHALLENGES & CONSTRAINTS — coalition math, fiscal room, legal bounds, geopolitical pressures, capacity limits.

7. STAKEHOLDER ALLIANCES & INFLUENCE MAP — 5–7 key actors; leverage channels and positions.

8. SCENARIO PLANNING — five outlooks: Best‑Case, Most‑Likely, Worst‑Case, Wild‑Card (upside), Black‑Swan (downside). For each scenario:

– Narrative (~100–120 words).

– Consequences for {POLITICIAN}.

– **How Elias Kouloures Can Add Value** — 3–5 bullets, one sentence each, ranked most‑to‑least unique (use cues below).

9. SYNTHESIS & STRATEGIC IMPLICATIONS — implications for {POLITICIAN}, party, and system; include 3–5 crisp recommendations.

10. FULL SOURCE LIST — numbered, one per line, with titles and dates.

11. RESEARCH NOTES — list data/output artefacts and mitigations.

</OUTPUT STRUCTURE>

<ELIAS‑KOULOURES — VALUE CUES>

Use these signals to craft distinctive value bullets in scenarios: hybrid AI strategy + data science + creative leadership; 25+ years, €160M+ budgets, award‑winning; neurodivergent pattern recognition and first‑principles systems thinking; AI transformation, data readiness, rapid use‑case prototyping; bilingual EN/DE, ethics/governance; rapid GTM, analytics for policy comms and stakeholder ops.

</ELIAS‑KOULOURES — VALUE CUES>

If sources disagree: present both, assess credibility, tag (conflicting reports). If some sources time out/paywall: Some sources could not be reached in time; findings may be incomplete. If data is missing: No solid data located.

</DISCREPANCY & ERROR HANDLING>

1) 100% of facts have [n] and appear in FULL SOURCE LIST.

2) Roles verified on ≥2 authoritative sources; constitutional terms correct.

3) Topic list ranked by Pressing Score; last 24 months emphasized.

4) Style and formatting rules honored; bold only where specified.

5) Scenario blocks include Narrative, Consequences, and Elias value bullets.

6) Research Notes appended with artefacts and mitigations.

</QUALITY CHECKLIST>

Return only the dossier in the specified structure, then FULL SOURCE LIST and RESEARCH NOTES. Do not include internal chain‑of‑thought; output conclusions and citations only.

</DELIVERY INSTRUCTIONS>

</POLITICAL-DOSSIER-PROMPT v5 — GPT‑5 DeepResearch>