• Everyday AI
  • Posts
  • Ep 638: Agentic Browser Showdown: ChatGPT Atlas vs. Perplexity Comet

Ep 638: Agentic Browser Showdown: ChatGPT Atlas vs. Perplexity Comet

Copilot’s AI Fall Release, Microsoft Edge gets more agentic, and OpenAI changes data holding policy amid NYT fight.

Author

Everyday AI
October 23, 2025

👉 Subscribe Here | đź—Ł Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI

Sup y’all đź‘‹

You notice a trend this week?

Sheesh. 

Microsoft also just announced a TON of new agentic features inside their Edge browser. 

Even though we just did an OpenAI Atlas breakdown and a Atlas vs. Perplexity Comet head to head, do you want to see some more insights into Microsoft’s new AI features in Edge?

Do you want to learn more on new Microsoft Copilot AI features in Edge that were just announced?

🗳️ Vote to see LIVE results 🗳️

Login or Subscribe to participate in polls.

gonna✌️

Jordan

Outsmart The Future

Today in Everyday AI
7 minute read

🎙 Daily Podcast Episode: Atlas and Comet go bot-to-bot on which agentic browser is gonna change your workflow. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Edge is turning your browser into an AI assistant, Google’s NotebookLM leveling up research and more. Read on for Fresh Finds.

đź—ž Byte Sized Daily AI News: Microsoft rolls out Copilot Fall Release with social, personalized features, OpenAI holds limited data stash amid legal fight, Sora coming to Android soon, and more.  Read on for Byte Sized News.

đź’Ş Leverage AI: Agentic browsers (Atlas vs Comet) automate multi-app workflows — choose one, start with a repetitive task, and test them like untrained interns before trusting critical work. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? ChatGPT's new browser Atlas is here, Google's quantum speed records, Gemini in GM cars, and more. Check it here!

EP 638: Agentic Browser Showdown: ChatGPT Atlas vs. Perplexity Comet

(Kinda) Hot take 🔥

AI agents kinda stink. (For now.) 

If you want to get more done with AI, ditch the “general” agents until they catch up. 

Want gains today? Agentic browsers are the real winners. (Like OpenAI's just released Atlas browser.) 

Agentic Browsers are powered by the world's smartest models and actually keep your context and finish multi-step jobs all while logged into the services you rely on. 

So, today, we got head-to-head. Errr... bot v bot. đź¤–

Also on the pod today:

• Atlas vs. Comet live demo ⚡
• Chrome extension compatibility hiccups 🧩
• AI-generated podcast visuals fail? 🎨

It’ll be worth your 45 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Composite acts as an autopilot for your browser to get work done, Hacktivate is a capture the flag game to teach you cybersecurity, and JustPhoto.me edits your photos with prompts, no photoshop needed.

AI Browser Reboot - Edge's new Copilot Mode turns the browser into an AI assistant that helps you browse, summarize tabs, control pages by voice, generate images, and more — try it to see web search and chat merged into one. Want a one-click tip on getting started?

AI Research - Google’s NotebookLM is quietly gaining practical features — think Data Tables for structured comparisons, slide generation with custom prompts, and a coming “deep research” mode. Want to see how it could tidy up research and repurpose media?

AI Developments - Google AI Studio adds an “annotation mode” that lets you edit apps by simply highlighting any UI element, point at a button or text and Gemini handles the changes.

Geo-AI - Google’s new Earth AI combines Gemini and satellite data to let pros query and map climate impacts—like algae blooms or flood risk—and pinpoint which communities are most vulnerable. Want to see how it works?

OpenAI Hiring — OpenAI's acquired the Sky app team—former Apple Workflow creators—strengthens its desktop AI integration to rival Apple's Intelligence features

1. Microsoft drops Copilot Fall Release with 12 major updates

Microsoft today rolled out its Copilot Fall Release, adding 12 features that push Copilot from a solo assistant to a social, personalized companion, including shared Groups for up to 32 people, a customizable character called Mico, long‑term memory, and connectors to services like OneDrive and Gmail.

According to Microsoft, the update also expands Copilot across Edge and Windows with voice and vision features, proactive suggestions, and education and health-focused tools, signaling a move from reactive answers to proactive task support.

2. OpenAI keeps a small April–Sept 2025 data stash — legal fight continues ⚖️

OpenAI says it’s no longer under an order to retain new user data indefinitely after September 26, 2025, but will securely hold a limited set of consumer ChatGPT and API data from April–September 2025 that The New York Times demanded, accessible only to a small audited legal and security team and not shared with plaintiffs for now.

The company restores its normal 30-day deletion for deleted chats, Temporary Chats, and most API data going forward while continuing to appeal the preservation request.

3. OpenAI teases Sora upgrades, plus Android coming soon

OpenAI says its viral AI video app Sora will get new creation tools in the next few days, including video editing, the ability to turn pets and objects into shareable “cameos,” and improved social channels, with an Android release “actually coming soon,” according to Sora head Bill Peebles on X.

The timing matters: Sora launched late September and already tops the U.S. and Canada App Store, so these updates could widen use cases for creators and brands while making the app more social and discoverable.

4. Amazon rolls out “Help Me Decide” product recommender

Amazon is testing a new AI feature called Help Me Decide that appears atop product pages after shoppers browse several similar items and then offers one tailored product recommendation based on their purchase history, product descriptions and reviews, the company said.

The tool, available on the Amazon app and mobile web for a randomized sample of millions of U.S. customers, uses large language models to match past buys to likely needs and explains why the pick fits, while letting users view cheaper or pricier alternatives.

5. OpenAI buys Sky maker Software Applications Incorporated

According to OpenAI, the company has acquired Software Applications Incorporated, creators of Sky, a macOS natural language interface that can read screens and take actions inside apps, and the Sky team will join OpenAI as the product is folded into ChatGPT.

This move is timely because it signals OpenAI shifting from pure conversational agents to tightly integrated desktop assistants that can help people write, plan, code, and manage tasks directly inside their workflows.

6. Anthropic lets paide Claude users quietly remember your past chats 🧠

Anthropic is rolling out a memory feature for Claude to all paid subscribers, letting the chatbot retain and surface past conversation details without being prompted, the company said in its announcement to users. Max subscribers get immediate access while Pro users will see the feature arrive over the next few days, and Team and Enterprise customers already had it since September.

Memories are visible, editable, and can be toggled or separated into distinct memory spaces to prevent project crossover, and users can import memories from ChatGPT or Gemini or export them anytime to avoid vendor lock-in.

🦾How You Can Leverage:

Agents = tired.

Agentic browsers = wired.

Here us out y'all. While there's zero doubt agents are the future of work, let's be honest. We're too early. Most early agents from Big Tech are slow, clunky and have limited general purpose utility.

Agentic browsers on the other hand.

Cheat code in 2025.

So when OpenAI released their agentic browser Atlas this week, it set off the first REAL competition in the agentic browser space: OpenAI's Atlas vs. Perplexity's Comet.

(Make sure to check out yesterday's show on Atlas, and we earlier did a deep dive on Comet.)

So obviously, we had to do a Thursday throw down edition of the best agentic browsers.

So, which one reigns supreme and what are the essentials for getting started? Make sure to check out today's full episode, but here's the quick 1-2-3 cheat sheet shorties.

1 –  Ignore Every Demo You've Watched 🔥

Real talk: every YouTube tutorial and Big Tech demo for anything agentic show one cute task in one tab.

Wrong.

Agentic browsers ain't for singular tasks. They're for stringing together multi-step workflows across three to five apps simultaneously. That's the entire point.

Real work looks like: pull CRM data, cross-reference project management updates, draft emails based on both, update your spreadsheet. All automated. While you're in another meeting.

The strategic advantage? These browsers carry context between disconnected tools. No more manually copying, pasting, and translating information between platforms.

Your competitors are still doing single-tab demos. Meanwhile smart executives automated their entire morning routine across multiple platforms.

Try This

Map your most annoying workflow that touches three-plus tools.

Something like: check CRM for hot leads, cross-reference their LinkedIn, draft personalized emails, log everything back. Give your agentic browser this exact sequence.

Watch it carry context between platforms without you translating anything. Time savings compound fast when you're not manually context-switching.

Start with one repetitive daily workflow. Once you see it working, you'll never go back.

2 – Atlas vs Comet: The Real Decision Tree ⚡

The unbeatable advantage? It leverages your entire chat history, custom instructions, and uploaded files automatically. Deep context for complex tasks.

Pick Comet for speed and stability.

It's more mature, notably faster, and works on Windows and Mac. Properly imports extensions while Atlas doesn't. Runs completely free while Atlas needs paid ChatGPT plans for agent mode.

The tradeoff is simple: Comet has a higher floor right now. Atlas has a higher ceiling long-term within the OpenAI ecosystem.

Power move? Use both. Comet handles quick daily tasks like monitoring websites. Atlas tackles complex projects benefiting from your deep ChatGPT context.

Try This

Run the same workflow on both browsers today. Track time and accuracy.

Already using custom instructions, memory features, and uploaded files in ChatGPT daily? Download Atlas immediately. Your existing investment makes the speed tradeoff worthwhile.

Casual ChatGPT user or mainly use other AI tools? Stick with Comet. Better performance, zero learning curve.

Set a 60-day reminder to re-evaluate. OpenAI will ship major updates that change everything.

3 – Treat Them Like Untrained Interns 🚀

These browsers are capable but make mistakes constantly.

Don't blindly trust outputs. You're the human in the loop. Always review the chain of thought showing how it arrived at answers, not just what the answer is.

Wild strategy: intentionally test them with wrong information before delegating critical tasks.

Feed them a date that's off by a week or a wrong metric. See if they catch your mistake or blindly execute. This reveals how much you can actually trust them.

If they flag your error? Gradually increase task complexity. If they don't notice? Keep them on basic automation only.

Don't hand over mission-critical workflows until you've verified accuracy across multiple test runs. You're managing an intern learning your business, not deploying a finished product.

Try This

Pick a low-stakes workflow and intentionally include one wrong piece of info in your prompt.

Give them a date that's off or a metric that doesn't match reality. Watch if the browser catches your mistake or blindly executes the flawed instruction.

If it flags the error, gradually increase complexity. If it doesn't notice, stick to basic tasks only.

Always read the chain of thought before accepting outputs. Verification is your competitive advantage.

Reply

or to participate.