The lowest price per tote cost $51 more on a 500-tote order.
Price per tote looks like the answer
Three fictional quotes arrive in three formats: $54 per case of 24 (Northfield), $2.10 each (Harbor) and $1.95 each (Pinecrest). Per tote, Pinecrest looks cheapest.
Source: Our test, fictional quotes, Oct 2026
The total has four moving parts
What you really pay depends on how many you must buy, the price, setup fees and shipping. Case sizes, minimums and shipping caps all change the total.
Ask for the total, not the price
Tested in GPT-5.5 (OpenAI API), Oct 2026
Give the AI your real quantity and ask for the working and every assumption. Paste the three quotes under the prompt.
Try it: Swap in your own quantity before you paste the quotes.
It rounds up and shows its steps
Real output · GPT-5.5 (OpenAI API) · Oct 2026
It rounded 500 up to 21 full cases, or 504 totes, then added $60 shipping because $1,134 is under the $1,200 free-shipping line.
Source: Our test, fictional quotes, Oct 2026
The hand check matched all three
I rechecked by hand: 21 × $54 + $60 = $1,194. Pinecrest is 500 × $1.95 + $150 setup + $120 capped shipping = $1,245. Harbor's 600 totes come to $1,340.
Source: Our test, fictional quotes, no tax
The math was right. The assumptions weren't settled.
Real output · GPT-5.5 (OpenAI API) · Oct 2026
The quotes don't settle these, so confirm them with the supplier. The short prompt also picked Northfield, but listed no assumptions.
Try it on your next quotes
Paste your next quotes, ask for the working and assumptions. Remove names, account numbers and other personal details first, and check the tool's data policy or your employer's approved tools before uploading real documents.
Try it: Recheck one row with a calculator before you order.
Sources and assumptions
- Our test: engine/ai-test.js runs quotes-before, quotes-after, quotes-before-2 (GPT-5.5, OpenAI API), recorded in record.tests: The real outputs on slides 5 and 7: GPT-5.5 put Northfield at $1,194, Pinecrest at $1,245 and Harbor at $1,340 in all three runs, named Northfield cheapest, and listed 7 assumptions in the quotes-after run. (Placeholder URL: the evidence is the recorded test, not a web page.) (checked 2026-10-09)
- Anthropic, Prompting best practices (Claude Platform Docs): Context only, NOT a slide claim: the current docs call manual chain-of-thought a fallback for when thinking is off, and say Claude Sonnet 5.5/Opus 5.5 should rely on built-in thinking. A prompt that asks the model to write out its reasoning 'may be declined'. So the post must not say Anthropic recommends 'show your working' for accuracy. Here 'show the working' is for the human checker, and the tested tool is GPT-5.5. The old long-context link now redirects here. (checked 2026-10-09)
Assumptions:
- All three quotes are fictional (Northfield Supply, Harbor Goods, Pinecrest Trading); prices invented for the test.
- Need = at least 500 totes; each supplier's smallest allowed quantity of 500 or more is bought (Northfield 504 via 21 full cases, Harbor 600 minimum, Pinecrest 500).
- Northfield's 'free on orders over $1,200' is read against product cost ($1,134), so $60 shipping applies; the model stated this assumption itself. Under any reading that adds only $60 shipping, the order is still under $1,200.
- No sales tax, rush, artwork or other fees, since none appear in the quotes.
- Per-tote headline prices: Northfield $54/24 = $2.25, Harbor $2.10, Pinecrest $1.95.
- Tested only in GPT-5.5 via the OpenAI API (label 'GPT-5.5 (OpenAI API), Oct 2026'), not the ChatGPT app. Claude runs failed to execute in this environment, so no slide may claim the prompt was tested in Claude.
- The vague prompt also got the right answer in both runs; the post must not claim it failed.
- Privacy line on the Try-it slide: remove names, account numbers and other personal details first, and check the tool's data policy or your employer's approved tools before uploading real documents.
The short version
- Price per unit can hide the real total.
- Ask for the total at your quantity.
- Ask for the working and assumptions.
- Recheck one row by hand.


