Opus 4.8 Read 3,000 Listings — Found 5 Niches I Can Win

About this video

RJ Martinez tests Anthropic’s Claude Opus 4.8 on his weekly print-on-demand niche research instead of coding or trading. He cites three relevant traits: it scored 84% on the Online Mind2Web browser benchmark, it was trained to flag uncertainty rather than invent keywords, and Claude’s effort control can be set to high or max — he uses max, despite the warning it may use excessive tokens. His inputs are an EverBee export of top-selling Etsy POD products (which he describes typing a keyword like ‘shirt’, ‘golfing shirt’ or ‘hiking shirt’ into Product Analytics and exporting) plus screenshots of the first 49 or 50 Amazon Hot New Releases designs. His prompt asks for five niches with low competition, real sales velocity and no oversaturation, each with design angle, seasonal hook and audience, showing its reasoning. After processing what he says were 3,000 listings, and filtering out designs that could get a Merch or Etsy account suspended for IP reasons, it returned: cottagecore silly goose/frog/mushroom garden; allied health pride beyond nurses — lab tech, ultrasound/sonographer, therapy, surg tech, with Lab Week as a hook; feminine preppy outdoors cute angler fishing, camping and lake; patchwork colorful floral personalized grandma names like Mimi, Gigi and Nana; and Christian parent faith-aesthetic ‘kingdom work’ motherhood/fatherhood boho floral verse tees — noting pickleball was essentially absent from the export. He then has it cross-check the niches against Amazon, and for niche three generate five Ideogram prompts plus an Etsy title, description and 13 tags, and an Amazon Merch brand, title, bullets and description. He stresses one of the five niches proved more competitive on Etsy than estimated, so he still verifies by hand: EverBee supplies the data, Opus makes sense of it faster.

Key takeaways

• Combine data sources: export EverBee Product Analytics for Etsy POD and screenshot the first 49-50 Amazon Hot New Releases, then feed both to the model together
• Run Opus 4.8 on max effort for this task and prompt for exactly five low-competition, high-velocity niches with design angle, seasonal hook, audience and verifiable reasoning
• Make the model filter for trademark/IP risk first — it flagged designs that could get a Merch by Amazon or Etsy account suspended
• Chain the follow-up: cross-check niches across Etsy and Amazon, then generate Ideogram prompts and a full listing (Etsy title, description and 13 tags, or a Merch listing) for one chosen niche
• Do not trust the output blindly: the creator found one niche’s competition was higher than estimated on Etsy and says Opus does not replace EverBee or manual platform checks

Episode details: published June 1, 2026 · 12:49 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are RJ Martinez’s own claims, shared for education — they are not typical or guaranteed.

← Previous: This Etsy Store Made $800K — I Cloned It With AI in 15 Min · Next: One Kittl Prompt = A Print-Ready POD Design… But Everyone’s Picking the Wrong AI Mode →

All RJ Martinez videos · RJ Martinez profile · RJ Martinez on YouTube