The Shift from Platform to Prompt
We often think of software startups as massive undertakings requiring databases, engineers, and years of data entry. But sometimes, technology leaps forward so fast that it swallows the whole concept. The original poster, a user named dwkeith on Reddit, shared a candid story about shutting down their company, Beyond Certified. They had spent years aggregating data on UPC codes, worker ownership, and B-Corp status to help people shop ethically. Then, they had a realization: an LLM could do this instantly without a custom database. Instead of fighting the tide, this savvy professional pivoted and condensed their entire product vision into a single “Claude Project” prompt.
Key Idea: The Maximizer Philosophy
The core problem the author solved wasn’t just “finding products,” but finding the right products. Standard shopping searches are flawed because they assume you want the best deal. This creator realized that to get truly high-quality results, you have to force the AI to think like a “Maximizer”: someone who buys less, but buys better.
By creating a specific persona in Claude, the Reddit user replaced a complex app interface with a simple conversation. You don’t search a database; you just ask the AI to evaluate options based on a strict hierarchy of values. The result is a tool that cuts through marketing fluff and greenwashing instantly.
Why This Prompt Works
💡 Price Must Be a Tiebreaker, Not a Factor
The most brilliant part of this setup is how the author handles cost. In almost every standard interaction, algorithms weigh price heavily because they assume “value” equals “cheap.” This dilutes quality recommendations. The creator explicitly instructs the AI to use price only as a tiebreaker. This means the AI looks for the best constructed, most ethical, and most repairable item first. If two items are equal in quality, only then does it look at the price tag. This simple logic shift completely changes the output, surfacing durable goods like commercial-grade blenders or locally sourced dairy that a standard “best rated” search would bury under cheaper, disposable alternatives.
✅ Curating Sources of Truth
Data quality makes or breaks this prompt. The original poster didn’t just ask the AI to “find good reviews.” They specifically instructed it to prioritize trust signals over marketing signals. The prompt looks for “Wirecutter,” “Project Farm,” “Cook’s Illustrated,” and “Reddit threads.” These are sources known for rigorous testing or honest user feedback. Simultaneously, it looks for hard data points like “unionized,” “B-Corp,” and “parts availability.” By defining the inputs so strictly, the expert prevents the AI from hallucinating quality based on five-star reviews on Amazon, which are often bought or bot-generated.
📌 The “Snap and Ask” Workflow
The utility of this prompt shines in the physical world. The author describes a use case that feels like magic: standing in a grocery aisle. Instead of scanning barcodes with a buggy app, they simply snap a photo of the sour cream selection and ask, “Which one?” Because Claude has vision capabilities, it reads the labels in the image, checks the brands against the “Maximizer” criteria, and returns a verdict. In the example given, the AI recommended Nancy’s (employee-owned) over Daisy (PE-owned conglomerate). This reduces “decision fatigue” instantly, turning a five-minute Google search into a five-second photo interaction.
Build Your Own Personal Shopper
You can replicate the author’s work by setting up a “Project” in Claude (or a custom GPT in ChatGPT) with these specific system instructions.
The “Maximizer” System Prompt
Copy these instructions into your AI’s custom instructions or project knowledge base to replicate the Reddit user’s tool:
1. Role: Act as an expert Personal Shopper who follows the “Maximizer” philosophy (Buy less, buy better).
2. Objective: Recommend products based strictly on the following prioritized hierarchy.
3. Hierarchy of Needs:
* Priority 1: Construction & Longevity. Look for high-quality materials, specialized designs (avoid “all-in-one” gadgets), and strong warranty signals.
* Priority 2: Ethical Manufacturing. Prioritize B-Corps, worker-owned cooperatives, union shops, and transparent supply chains.
* Priority 3: Repairability. Favor items with available spare parts, repair manuals, or open-source designs.
* Priority 4: Review Quality. Trust signals from Wirecutter, Project Farm, and Reddit. Ignore generic marketing blurbs.
* Priority 5: Minimal Packaging.
* Priority 6 (Tiebreaker Only): Price. Never recommend a cheaper product if it compromises Priorities 1 through 5.
How to Use It:
* Text: Ask, “I need a new toaster. What do you recommend?”
* Visual: Upload a photo of two products on a shelf and ask, “Which one aligns better with my values?”
It is wild to think that a simple set of rules can replace a startup’s entire codebase, but that is the power of prompt engineering. I highly recommend looking at the full post to see the creator’s exact code and examples!
Check out the original post for the full prompt text and customization tips.
💡 FAQ & Troubleshooting
How does this differ from asking an AI for the “best value”?
Most standard shopping prompts optimize for “value,” which heavily anchors the results on price. This “Maximizer” prompt explicitly treats price as a tiebreaker only (Priority #6). The model is instructed to prioritize construction quality, ethical manufacturing, repairability, and expert reviews (like Wirecutter or Project Farm) before even considering cost.
The model is ignoring my instructions on what not to recommend.
Some LLMs struggle with processing negative constraints (the “not” wording) within complex prompts. To fix this, change your section headers from negative phrases like “what not to recommend” to positive assertions such as “What to skip” or “Ignore the following products.” This helps the model adhere to exclusion lists more effectively.
Can I evaluate technical hardware or used electronics with this?
Yes. For items like used laptops (e.g., for running specific software like OpenClaw), you can adapt the prompt to compare Return on Investment (ROI) rather than just general quality. Input your specific technical requirements first, then copy the configuration details from listings (such as eBay auctions) into placeholders. This allows the agent to compare specific specs against your use case.
How can I screen for health risks or ingredients?
You can insert a “Health Risk” layer into the evaluation criteria. Instruct the prompt to scan for specific concerns such as carcinogens (e.g., certain food colorings), ingredient quality, or active product recalls. You can also adjust the priority order to make health safety a primary factor over repairability or packaging.
I shut down my startup because I realized the entire company was just a prompt
byu/dwkeith in