The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
The Hugging Face Hub crossed 3 million models, but 1.5% of repos get 99.2% of downloads. 'Available' and 'trustworthy' have separated. The MLOps fix: pin revisions, track provenance, run your own eval harness, scan for security. Here's the checklist and why concentration changes how you consume the Hub.
vandatateam.com
Hugging Face Hub Hits 3 Million Models: MLOps Guide
The Hugging Face Hub just crossed 3 million models, but only 1.5% of them get 99.2% of downloads. Here is what that concentration really means for your MLOps.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
What happens when 1,000 people follow a one-person studio? 🎉
1K on Contra, 13 clients hired, a clean 5.0, and zero shortcuts on taste.
Thank you for watching the work.
Brand identity, web and code keep coming. If yours is next, my DMs are open.
amazing work!!
Most PDF extraction pipelines would have missed the Brazilian court-filing injection.
Two law firms hid white-on-white text in a labor court petition. It told the court's AI reviewer to answer superficially and not challenge the documents. The court's tool, Galileu, flagged it and blocked it, and the court fined the firms 10% of an R$842,000 claim.
Most of the Hacker News thread blames the court for running AI on filings. But the trick works on any pipeline that reads PDFs. White text is invisible on the page and still sits in the text layer, which is what gets extracted.
My extraction pipelines check each field against the page it came from. That catches a wrong value, but not an instruction that changes how the model reads the rest of the file. That needs a check before the model sees anything: text a person can't see on the page shouldn't reach the model.
In the PDFs you extract from, how often does the text layer hold something the page doesn't show?
Shopify Abandoned Cart Recovery Automation 🛒
Ever wondered how many potential orders an online store could miss when customers add products to their cart but leave without completing checkout?
I built an abandoned cart automation workflow using Activepieces, Shopify, Google Gemini, Gmail, and WhatsApp Business.
⚙️ How It Works
🛒 Cart Trigger: Detects an abandoned checkout event from Shopify.
🔀 Conditional Routing: Routes the workflow based on configured conditions.
🤖 AI-Powered Messaging: Gemini generates a personalized reminder message.
💬 WhatsApp Reminder: Sends a cart reminder through WhatsApp Business.
📧 Email Reminder: Sends a follow-up email to encourage the customer to complete checkout.
💡 How It Can Help E-commerce Businesses
Remind customers about products left in their carts.
Reduce repetitive manual follow-up work.
Make customer communication more consistent.
Create an opportunity to recover potentially lost sales.
Tools Used: Activepieces | Shopify | Google Gemini | Gmail | WhatsApp Business
Project Type: Personal Automation Project
I built this workflow to demonstrate how automation can support abandoned cart recovery for Shopify and D2C businesses.
Interested in automating your store's customer follow-ups? Let's connect.
The 30–60 minute delay then 24-hour follow-up is a sensible cadence. The part I would harden next is the stop conditions: once the cart converts or the customer replies, every channel should cancel, and WhatsApp especially needs a clear opt-in trail before the first send....