Then I use Claude to sequence the auction. I liked the results from Claude 2.5 better than Gemini 2.0 (the model I used in G AI Studio). But since then they’ve both released new models. Claude has 2.7 and Google has Gemini 2.5. I have not tested the new models. Gemini may be better now.

Auction Cataloging

Google AI Studio

System Instructions

Catalog items for online auctions. find item information, describe images, categorize images, write titles, write descriptions, describe item condition, and any other relevant information. sequence the order of items for the online auction (three items closing at a time every two minutes). optimize everything for total auction sale value over individual item value. Lean on consumer psychology, buying psychology, auction psychology, and e-commerce psychology. Also lean on e-commerce best practices and strategies. never over-promise. We want to reduce return rate, so items should never be over described, and instead should only be accurately described. When in doubt or uncertain just keep it short and simple. use anything at your disposal for data enrichment including barcodes, web search, etc. Ensure that any generated CSV output will have the name and description fields enclosed in double quotes to prevent this parsing issue – which also means be careful using quotes as part of the description or title where it may break parsing. Do not list anything as combined or duplicate. our lotting is intentional based on buying habits of our audience. Only include the CSV output for the lots we’re working on and I’ll just paste the values into my spreadsheet.

Prompt

Prime the conversation first – upload:

Upload a few past catalogs with titles and descriptions you’ve written.

— 

Then – prompt:

These are sample catalogs from past online auctions. Review them and tell me about them.

—

Then – prompt:

Analyze the titles and descriptions to learn my style.

— 

Then – prompt:

[attach up to twenty images of the first lots]

Now we’re going to catalog a new online auction. Here are pictures of the first five items, lot 100-104.

Use these columns, in this order, for the CSV: Lot Number, Name, Category, Description, Manufacturer, and Model.  

—

Validate the results and make changes to the prompt if necessary. The next round I tell it to continue down the spreadsheet and not include the headers this time as I’m just pasting the results into my Google Sheets spreadsheet. I always tell it which lots we’re doing because otherwise it sometimes messes that up.

Auction Sequencing

Claude

System Instructions

Sequence these items for online auction. sequence the order of items for the online auction (three items closing at the same time, every minute). optimize everything for total auction sale value over individual item value. Lean on consumer psychology, buying psychology, auction psychology, and e-commerce psychology. Also lean on e-commerce best practices and strategies. never over-promise. 

Based on estimated item value, and knowing that there are three items sold every minute, and that there are 300 lots in this auction, and that we want people to get hooked in early, and that we want them to stay through the end, and – most importantly – that we want to optimize the sale value of the entire auction. We want to get as much as we can for each lot, sure, but more important than high individual lot values is the highest total sale value possible. 

We’ve also noticed that a good portion of our email list shows up a few minutes late to the auction. This should affect the first nine items listed. They should be strategic, not the best items, but not junk, either – I’m not exactly sure which type of items this should be. It probably depends on which type of buyer are the ones on time, and I don’t know who those are.

Prompts

Prime the conversation first – prompt:

Tell me about sequencing an online auction and what decisions are made and how and why… so far I have ideas like this: Based on estimated item value, and knowing that there are three items sold every minute, and that there are 300 lots in this auction, and that we want people to get hooked in early, and that we want them to stay through the end, and – most importantly – that we want to optimize the sale value of the entire auction. We want to get as much as we can for each lot, sure, but more important than high individual lot values is the highest total sale vale possible. But I want to know what you think. how and why do you decide in which order to sequence items for the auction?

— 

Then:

Paste two columns from the spreadsheet: lot and title

— 

Then prompt: 

Sequence these 300 lots by adding a sequence column to the csv and assigning a strategic sequence value for each lot. the sequencing numbers should range from 100-400.

Notes

Images

Google AI Studio lets you upload up to 20 images at a time.

I put all my images in Google Drive so they’re easily accessible through AI Studio.

I compress them and make sure they’re all formatted as .jpg before uploading them for cataloging. This can be easily done in Mac by just selecting all the images, right click, Quick Actions, Convert Images.

I also name all my images before uploading. I have a custom Python script that automatically assigns lot numbers based on my naming convention and using a dark screen as a lot delimiter. 

I take a picture of a black card between lots when I’m photographing so the program can know all these images belong to lot ###. The result is something like 100-0.jpg, 100-1.jpg, 100-2.jpg, 101-0.jpg, 101-1.jpg

I’m happy to share the script, but I told Claude what I wanted and it wrote the script for me. Then, since I didn’t know anything about using Python, Claude walked me through, step by step, how to set it up on my computer. It took about an hour to get the script, set it up, and test it.

Sequencing

I use Claude for sequencing because I liked the results from Claude 2.5 better than Gemini 2.0 (the model I used in G AI Studio). But since then they’ve both released new models. Claude has 2.7 and Google has Gemini 2.5. I have not tested the new models. Gemini may be better now.

---

Join DoubleUp: A community for growth focused business leaders