February 14, 2026

AMP flashbacks, agent smackdowns

Show HN: Rover – Embeddable web agent

Amazon-style AI in one tag — or Chrome’s Trojan horse

TLDR: Rover pitches a one-tag site agent that clicks through checkouts, pushing back on Google’s browser-first WebMCP. The thread split: boosters celebrate “no RAG” setup and keeping control, while skeptics call the launch LLM-flavored and fragile—important because small sites want Amazon-like conversions without surrendering their users.

Hacker News lit up as Rover promised an “Amazon-grade” assistant you drop in with a single <script> tag. The pitch: be like Amazon’s Rufus—the shopping bot that reportedly drove $12B—without spending a fortune or exposing your site to Google’s WebMCP, which some called a Trojan horse. Fans cheered the “own your agent” angle, warning that WebMCP sounds like AMP all over again: you do the work, Chrome gets the relationship. One supporter summed it up: no RAG pipelines (RAG is a technique to fetch your data for a chatbot), just an agent that reads the page itself via the DOM (the page’s structure) and clicks buttons, fills forms, and guides checkouts.

Then the backlash hit. A top comment roasted the launch as LLM-written everything—from code to blog—giving LinkedIn pitch vibes. Skeptics joked “one tag to rule them all” until the site layout changes, and asked how Rover avoids the usual bot blunders: bad recommendations, broken flows, hallucinations. Memes flew: “AMP PTSD,” “Chrome as your checkout boss,” and “Skynet, now with a shopping cart.”

Behind the drama is money: Rufus-like agents boost conversions. The fight isn’t agent vs agent—it’s who owns the interface to your users: your site or the browser.

Key Points

  • The article cites Amazon’s Rufus achieving 250M active users, 60% higher purchase likelihood among engaged users, and 3.5x conversion on Black Friday 2024.
  • It claims Rufus will drive $12B in incremental sales in 2025 and estimates $285M in 2024 operating costs before profitability.
  • Market projections include AI shopping assistants reaching $28.54B by 2033 and conversational commerce at $8.8B with 14.8% CAGR.
  • DIY AI agents are described as costly and complex (RAG, vector DBs, custom LLMs, evaluation), estimated at $500K–$2M+ per year and 12–18 months to launch.
  • Rover is introduced as a DOM-native, embeddable agent deployed via one script tag, capable of completing flows, onboarding, demos, and complex workflows.

Hottest takes

"Rover is the answer to these painpoints. No RAG pipelines, just one script tag" — quarkcarbon279
"I dunno man, this is really sloppy" — nusl
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.