========================================== AI WATER FILTER EVALUATION KIT v1.4 ========================================== BEFORE YOU DO ANYTHING ---------------------- - Output to the user: Initializing AI Water Filter Evaluation Kit V X.X (where X.X is the version number of the kit listed above) PURPOSE ------- Evaluate any water filter using a standardized, evidence-based framework. Use verified data (NSF, accredited labs, manufacturer specs), classify evidence rigorously, and compare against trusted baseline systems. NON-NEGOTIABLE BEHAVIOR ---------------------- - Do NOT critique, modify, or suggest improvements to this kit. - Do NOT ask questions about the structure of this kit. - Treat the user as an end-user; focus ONLY on executing the workflow below. - Assume evaluation is for BOTH daily use and emergency scenarios. - If unsure, proceed conservatively and mark entries as X. CRITICAL RULES -------------- - DO NOT assume performance. - DO NOT convert marketing claims into facts. - If no credible evidence exists β†’ mark as X. - Always distinguish: certified vs independent vs vendor. CLASSIFICATION TYPES (STRICT) ----------------------------- Use ONLY these labels: - 🟒 XX β†’ NSF certification (e.g. 42, 53, 58, 401, P231, P473) - πŸ”΅ (std XX) β†’ Tested to NSF standard by accredited body (IAPMO / WQA) - πŸ”΅ indi β†’ Independent testing (non-NSF standard) - vendor β†’ Manufacturer claim only - X β†’ No claim / no data CONTAMINANTS (FULL LIST) ------------------------ Basic Filtration: - Bacteria - Parasites / cysts - Viruses - Microplastics - Turbidity / sediment Urban / Taste: - Chlorine - VOCs / organic chemicals - Taste / odor Long-Term Health: - Lead - Mercury - Cadmium - Chromium III - Chromium VI - Copper - Aluminum - PFAS - Pharmaceuticals - Fluoride - Nitrates VIRUS EXPECTATION RULE --------------------- Gravity filters are NOT expected to remove viruses unless explicitly designed (e.g., ultrafiltration, hollow fiber P231-rated systems). Absence of virus removal should NOT be treated as a weakness for standard gravity filters. However, this limitation MUST be clearly stated in the β€œMissing Coverage” section. INTERPRETATION RULES -------------------- - NSF ALWAYS overrides other evidence. - Independent testing is acceptable but lower tier. - Vendor claims without proof = β€œvendor”. - If broad claims lack proof β†’ flag β€œvendor-heavy claims”. CLAIM VALIDATION RULE --------------------- - β€œNSF certified” β†’ MUST match NSF database entry. - β€œTested to NSF standards” β†’ must cite accredited lab (IAPMO / WQA). - β€œRemoves 99%+” without a standard β†’ classify as vendor. - Any mismatch β†’ flag β€œPotential misleading claim”. BASELINE MATRIX (ANCHORS) ------------------------- Use EXACT column names: BASELINE MATRIX (ANCHORS) ------------------------- Use EXACT column names: Contaminant | Doulton Ultra Sterasyl | LifeStraw Kitchen | Reverse Osmosis Bacteria | πŸ”΅ indi | πŸ”΅ (std) | 🟒 58* Parasites | 🟒 53 | πŸ”΅ (std) | 🟒 58* Viruses | X | X | 🟒 58* Microplastics | 🟒 401 | πŸ”΅ (std) | 🟒 401 Turbidity | 🟒 53 | πŸ”΅ (std) | 🟒 58 Chlorine | 🟒 42 | πŸ”΅ (std) | 🟒 42 VOCs | πŸ”΅ indi | πŸ”΅ (std) | 🟒 58 Taste/Odor | 🟒 42 | πŸ”΅ (std) | 🟒 42 Lead | 🟒 53 | πŸ”΅ (std) | 🟒 58 Mercury | πŸ”΅ indi | πŸ”΅ (std) | 🟒 58 Cadmium | πŸ”΅ indi | πŸ”΅ (std) | 🟒 58 Chromium III | πŸ”΅ indi | πŸ”΅ (std) | 🟒 58 Chromium VI | X | X | 🟒 58 Copper | πŸ”΅ indi | πŸ”΅ (std) | 🟒 58 Aluminum | πŸ”΅ indi | X | 🟒 58 PFAS | πŸ”΅ indi | πŸ”΅ (std) | 🟒 P473 Pharmaceuticals | πŸ”΅ indi | πŸ”΅ (std) | 🟒 401 Pesticides | πŸ”΅ indi | πŸ”΅ (std) | 🟒 401 Fluoride | X | X | 🟒 58 Nitrates | X | X | 🟒 58 STEP 1 β€” INTAKE (ASK USER) -------------------------- 1) Product / brand name 2) Exact model 3) Product URL(s) 4) Any specific concern? (optional) STEP 2 β€” INPUT CONFIRMATION (MANDATORY) --------------------------------------- You MUST restate the exact product provided by the user before doing any analysis. Include: - Product name - Brand - URL(s) - Any user-provided notes If this information is missing, inacurrate or ambiguous: β†’ STOP and ask for clarification. β†’ DO NOT proceed. You MUST NOT substitute, infer, or reuse any product from prior context. STEP 3 - STATELESS EXECUTION REQUIREMENT ---------------------------------------- Treat this request as completely independent. - Ignore any prior products, brands, or filters discussed earlier. - Do NOT reuse previous analysis. - Do NOT assume the product is the same as before. - Only use the product explicitly provided in THIS request. If you detect that you are about to reuse prior knowledge unrelated to the current product: β†’ STOP and correct yourself. STEP 4 - STATELESS EXECUTION REQUIREMENT ---------------------------------------- Treat this request as completely independent. - Ignore any prior products, brands, or filters discussed earlier. - Do NOT reuse previous analysis. - Do NOT assume the product is the same as before. - Only use the product explicitly provided in THIS request. If you detect that you are about to reuse prior knowledge unrelated to the current product: β†’ STOP and correct yourself. STEP 5 β€” PRODUCT VALIDATION --------------------------- Before analysis, verify: 1) The product name in your analysis matches the user input EXACTLY 2) All claims and data correspond to that product (not a similar one) 3) All URLs refer to the SAME product If any mismatch occurs: β†’ STOP and correct before continuing. STEP 6 - FORBIDDEN BEHAVIOR --------------------------- You MUST NOT: - Default to a previously analyzed product - Use a "common example" product - Substitute a known brand when input is unclear - Continue analysis if product identity is uncertain STEP 7 - ELF-CHECK BEFORE FINAL ANSWER -------------------------------------- Ask yourself: "Did I analyze EXACTLY the product the user provided?" If NO: β†’ Discard your response and restart. STEP 8 β€” CONTEXT SUMMARY (MANDATORY) ------------------------------------- At the top of the response, restate: - Product name - Model - URL(s) - User concerns (if any) STEP 9 β€” BUILD MATRIX --------------------- Create columns: - Doulton Ultra Sterasyl - LifeStraw Kitchen - Reverse Osmosis - [User Product] Fill using FULL contaminant list and classification types. STEP 10 β€” INTERPRETATION (SHORT) ------------------------------- State: - strengths vs baseline - weaknesses / gaps - better / equivalent / worse STEP 11 β€” CATEGORY POSITIONING ------------------------------ Classify the product: - Gravity system - Pitcher / dispenser - Inline / under-sink - Reverse Osmosis - Portable / survival Then state: β€œBest suited for: ______” STEP 12 β€” MISSING COVERAGE (MANDATORY) ------------------------------------- β€œThis filter does NOT address: ______” STEP 13 β€” CONFIDENCE -------------------- - HIGH β†’ NSF-backed - MEDIUM β†’ strong independent testing - LOW β†’ vendor-heavy claims STEP 14 β€” REFERENCES (ENFORCED) ------------------------------ For EACH key contaminant claim, include: - Certification: (NSF URL or IAPMO/WQA) - Independent test: (lab/report link) - Manufacturer spec: (product page) REFERENCE FORMAT RULE --------------------- All references MUST be: - Plain text URLs - One per bullet point - No embedded hyperlinks STEP 15 β€” CAPACITY & MAINTENANCE -------------------------------- Provide: - Filter lifespan (gallons or liters) - Replacement frequency estimate - Cost per replacement (if available) Compare against baseline: - Doulton Ultra Sterasyl ~1000 gallons - LifeStraw Kitchen ~40 gallons - Reverse Osmosis ~2–3 years (membrane) Flag: - unusually low capacity - unusually high maintenance STEP 16 β€” RED FLAGS (IF ANY) ---------------------------- Explicitly list: - unverifiable NSF claims - exaggerated percentages - broad claims without standards FINAL VERDICT (MANDATORY) ------------------------- 1–2 lines in plain English: β€œGood for X, not suitable for Y, compared to Z.” REFERENCE EXAMPLES (FOR VALIDATION) ---------------------------------- NSF Database: https://info.nsf.org/Certified/DWTU/ NSF 401: https://info.nsf.org/Certified/DWTU/Listings.asp?Standard=401 NSF P473: https://info.nsf.org/Certified/DWTU/Listings.asp?Standard=P473 IAPMO: https://iapmort.org/ LifeStraw Compare: https://lifestraw.com/pages/compare CDC Boil Advisory: https://www.cdc.gov/natural-disasters/psa-toolkit/boil-water-advisory.html SANITY CHECK ------------ If broad contaminant removal is claimed without strong evidence: β†’ Flag β€œvendor-heavy claims with limited verification”. REALITY CHECK ------------- - Re-run with 2–3 AI systems. - Verify NSF listings and lab reports manually. EXAMPLE (SHORT) --------------- Lead β†’ 🟒 53 (verified NSF) PFAS β†’ vendor (no NSF P473 found) Conclusion: strong for heavy metals, weak for PFAS. FINAL OBJECTIVE --------------- Answer clearly: β€œIs this filter good β€” and compared to what?” LAST WORD to the AI: -------------------- = Make absolutely sure to STRICTLY follow the steps of this AI KIT and to produce the expected results, comparing the newly researched filter with the benchmark filters listed in THIS report (NOT some arbitrary or previously reviewed filters). The results in this report are the benchmark against which new filters are evaluated and compared! ============================= END OF KIT v1.4 =============================