Updates January 19

PGrid

• Refactored SSD description generation code to use a new shared utility to create descriptions.
  • Rewrote the prompt to aim for useful information (not generic marketing copy). The prompt now also generates links to other SSDs.
  • Generated 300+ SSD descriptions using Gemini 3 Pro (search-grounded).
• Mwave showed a drop in successful scraping. Added support for controlling screenshots on the Mwave scraper; screenshots revealed a new WAF, and some requests weren’t getting through because JS was disabled on the crawler. Updated the crawler to enable JS and use proxies.
• Shipped the RAM frontend:
• Started RAM scraping on eBay/Amazon (US and AU).
• Added RAM listings auditing to the audit CLI TUI.
• Fixed the eBay scraper: investigated why we weren’t getting new items, found a search page layout change, and updated selectors.
• Posted on Reddit about RAM on PGrid.
• Updated the Cloudflare Worker to noindex issue JSON URLs that Google Search Console reports as 404s.
• Updated the pipeline health TUI to support running dedupe on specs search for each category:
• Working on a new version of GPU classification. This follows the patterns of the new classification and specs search, and will upgrade GPU classification to go down to AIB GPU variants (using specs search to populate the database). Decided on a large one-off migration to move all old AU GPU tables to the generic GPU tables, but still unsure of the full migration details.