AI ബ്രൌസർ War?

Perplexity - Comet Release cheyyunnu..

OpenAI Atlas erakkunu..

2 perum Agentic Browser market sharinu vendi adiyanu..

Google Chrome version 143 yil agentic capabilities kaanum enn puthiya leaks…

Chromil Agentic capabilities vannal,

Open AI kkum perplexity kkum pidichu nikkan padarikkum..

Enthayalum ivark 2 perkum athinu munne oru pani kititund..

OpenAI Atlas with Major Security Issue

Reddit Sues Perplexity

Ithu mathramalla vereyum interesting AI news ond..

I’m Alex, Welcome to L8R by Innov8.
Let’s DIVE DEEP 👨‍🏫👇

In today’s post:

• ChatGPT Atlas Can Be Tricked OpenAI Admits Real Risks

•  Reddit Sues Perplexity for “Industrial-Scale” Scraping

•  Google’s “Quantum Echoes” on Willow: 13,000× Faster

•  Amazon’s New Robot Army + AI Glasses, and Jassy’s Plain Talk on Jobs

ChatGPT Atlas Can Be Tricked OpenAI Admits Real Risks

OpenAI launched a new browser agent called ChatGPT Atlas

That can click around the web and do tasks for you.

But security experts and even OpenAI’s own team say it can still be fooled by “prompt injection”

Sneaky instructions hidden inside web pages that make the agent do the wrong thing.

🔍 Key Points:

• Atlas has an “agent mode” that controls the page for you; testers found the safety step called “Watch Mode” did not always turn on for sensitive sites like banking or GitHub.

• OpenAI’s CISO, Dane Stuckey, says prompt injection is not solved yet; misfires like wrong purchases or data leaks can still happen.

• OpenAI’s system card says Atlas must ask for user confirmation before sending money or messages, and it enables Watch Mode on sensitive sites; memory is disabled to reduce data leaks.

• Security writers warn attackers can even hide instructions in places like CSS or images, so users should supervise agent actions.

📰 What’s New?
OpenAI rolled out Atlas with strong promises, but the launch raised more questions about how it fights prompt injection. Testers highlighted gaps, and OpenAI pointed to its system card with mitigations like confirmations, spending limits, and Watch Mode.

⏭️ What’s Next?
Expect rapid patches, clearer rules, and more red-teaming; researchers will keep publishing new attack tricks, and OpenAI says it will keep tightening guardrails.

From our Partners

Find out why 100K+ engineers read The Code twice a week.

That engineer who always knows what's next? This is their secret.

Here's how you can get ahead too:

• Sign up for The Code - tech newsletter read by 100K+ engineers

• Get latest tech news, top research papers & resources

• Become 10X more valuable

Join 100k+ engineers

Reddit Sues Perplexity for Scraping

Reddit took Perplexity to court, saying the AI startup used data brokers and Google’s caches to suck up Reddit posts without permission.

Reddit says it ran a sting and caught Perplexity repeating a hidden test post.

🔍 Key Points:

• Lawsuit filed in New York federal court names Perplexity plus Oxylabs, AWMProxy, and SerpApi, accusing them of bypassing protections and scraping through Google results.

• Reddit says this is “industrial-scale” scraping that dodges its pay-for-data licensing; it already licenses to Google and OpenAI.

• Reports say Perplexity denies wrongdoing and plans to fight, arguing access to public knowledge.

• The case could set rules on whether taking content from search caches breaks the law or counts as fair use.

📰 What’s New?
The complaint describes a “sting” where Reddit made content only Google could see; Perplexity allegedly reproduced it, which Reddit says proves scraping. Multiple outlets picked up the story the same day the suit was filed.

⏭️ What’s Next?
Watch for motions on DMCA anti-circumvention and unfair-competition claims; more publishers may file similar cases as data licensing becomes a bigger business.

Google’s “Quantum Echoes” on Willow: 13,000× Faster

Google’s Quantum AI team says its Willow chip ran a new algorithm called “Quantum Echoes” much faster than the best supercomputer. and the result can be checked and repeated, which is a big deal.

🔍 Key Points:

• Google claims a verifiable quantum advantage: Quantum Echoes ran about 13,000× faster than classical methods on a top supercomputer.

• The algorithm runs a circuit forward, nudges one qubit, then runs backward to create an “echo” that reveals how information spreads.

• Team used Echoes as a “molecular ruler” with UC Berkeley, matching and extending NMR-style measurements on molecules.

• Experts warn advances like this make “harvest-now, decrypt-later” attacks more urgent; move to post-quantum crypto.

📰 What’s New?
Unlike past “speed-show” demos, this one is verifiable: another good quantum device can repeat the number it outputs. That helps turn quantum hype into real, testable progress.

⏭️ What’s Next?
Google is pushing toward a long-lived logical qubit and real applications within a few years; standards bodies and companies need to accelerate post-quantum encryption rollouts now.

💡Amazon’s New Robot Army + AI Glasses.

Amazon showed Blue Jay (a multi-arm robot station), Project Eluna (an AI “teammate” for managers), and smart delivery glasses for drivers.

CEO Andy Jassy also told employees that AI will reduce some corporate roles in the next few years. 🥀 

🔍 Key Points:

• Blue Jay collapses three warehouse steps into one station that can pick, stow, and consolidate in a tight space, built using AI and digital twins.

• Project Eluna is an “agentic AI” system that watches data feeds, predicts bottlenecks, and suggests staffing moves.

• Smart delivery glasses give drivers a heads-up display with scanning and turn-by-turn walking guidance; designed with input from hundreds of drivers.

• Jassy’s June memo: as AI and agents roll out, Amazon expects fewer people in some jobs and fewer total corporate roles over time.

📰 What’s New?
Alongside “one-million-plus robots” across sites, Amazon is layering a fleet brain (Eluna and other AI) and new driver wearables to push speed and safety. The leadership message is direct: AI brings efficiency, and team sizes will shift.

⏭️ What’s Next?
If Blue Jay and Eluna hit targets, expect wider rollouts into same-day sites, more AI in routing, and more debate about “cobots vs jobs” as automation scales.

🚨 Quick L8R Summary

• 🧠 ChatGPT Atlas: OpenAI’s new browser agent can still be tricked by “prompt injections” sneaky website code that makes it do things you didn’t ask for.

• 💥 Reddit vs Perplexity: Reddit sued Perplexity for secretly scraping posts through Google’s cache - a case that could rewrite AI data-use rules.

• ⚛️ Google Quantum Win: Google’s 105-qubit Willow chip ran its “Quantum Echoes” algorithm 13,000 × faster than supercomputers - and the result can actually be verified.

• 🤖 Amazon Automation: Amazon rolled out Blue Jay robots, Eluna AI, and smart AR glasses - while CEO Jassy warned some jobs will shrink as AI takes over.