WebAI StudioModeration Cascade
EmbeddingsThe production pattern for cheap moderation: embeddings clear the obvious 90% in microseconds, the LLM judges only the borderline.
Why On-Device?
Cascades are how real systems afford moderation — and on-device they cost literally nothing. Watch the counters: most comments never touch the language model.
Interactive Demo
This tutorial saved my weekend, thank you!
BUY CHEAP WATCHES >>> best-deals-watch dot com
You're an idiot and everyone here knows it.
Great write-up, bookmarked for later reference.
Make $5000/week from home, DM me now!!!
Could you do a follow-up on WebGPU support?
Only a complete moron would ship this garbage.
The dark mode on this site is gorgeous.
🔥🔥 FREE followers at insta-boost dot net 🔥🔥
I disagree with the benchmark methodology, but the data itself is useful.
Wow, genius idea. Really groundbreaking stuff. Slow clap.
I wrote a longer rebuttal on my blog if anyone is interested.
First time commenting — this community seems really helpful.
Implementation Code
Keep exploring
56 demos, all running on-device.
Cluster & Label
Group raw user feedback into themes with k-means over embeddings, then let the Prompt API name each cluster.
The Omnibox
One input, every API: whatever you paste is intent-routed to the right tool — translate, summarize, proofread, rewrite, or answer.
Photo Semantic Search
Drop in your photos, and search them by meaning — "food on a table", "someone smiling" — without one pixel leaving your device.