• Everyday AI
  • Posts
  • AI Agents: From automation to super agents, 10 AI Agents you should know

AI Agents: From automation to super agents, 10 AI Agents you should know

Meta’s AI glasses blunder, Google brings agentic features to Chrome, NVIDIA invests $5b in rival and more

Author

Everyday AI
September 18, 2025

Sup y’all 👋

Since yesterday you OVERWHELMINGLY said you wanted more AI agent content…. we delivered with today’s podcast episode.

And while we normally have a healthy dose of agent content from the big players (ChatGPT, Google, Microsoft and Anthropic) I was wondering…. what other AI agent content would you wanna see?

What other AI agents do you wanna see?

Vote to see results.

Login or Subscribe to participate in polls.

(Make sure to check the bonus content at the end of this newsletter.)

✌️

Jordan

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: There’s hundreds of AI agents. We gave you the Top 10. Give it a listen.

🕵️‍♂️ Fresh Finds: Deepseek makes more claims about cheap models, Perplexity’s new agent, Mistral’s new small model impresses and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Meta’s AI glasses blunder, Google brings agentic features to Chrome, NVIDIA invests $5b in rival and more.  Read on for Byte Sized News.

🧠 Learn & Leveraging AI: We caught through the AI Agent fluff and gave you the best use-cases for the Top 10 agents. So, what’s the takeaway? Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about Reddit talking bigger AI partnership with Google, Gemini and ChatGPT take home coding gold, Jensen Huang going bananas and more. Check it here!

AI Agents: From automation to super agents, 10 AI Agents you should know 🦾

There's hundreds of agents. 🤖

Most you should ignore.

These 10 though.... you've gotta know them.

Today, we went over Agents 101, how they work, and the 10 you should be paying most attention to.

Also on the pod today:

• 7 types of AI agents explained 🧩
• AI agents creating full apps 🚀
• Agent pitfalls: cost creep warning! ⚠️

It’ll be worth your 50 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 – Envelope is an AI agent for event planners, StackOverflow is making an AI-powered comeback, Sudio is a single API for all LLMs

AI Models — Deepseek claims that its new R1 model costs only $294,000 to train, which doesn’t count the costs of its underlying V3 foundation model. (We’re not buying it.)

AI Updates — Google announced that their Gems, personalized versions of Gemini, are finally shareable.

Perplexity Agent — Perplexity’s CEO confirmed they’re working on a server-side agent, not just a browser agent.

Small Models — Mistral’s Magistral 1.2 pairs small, local-ready models with stronger reasoning and new vision. Find out how they did it.

AI in Finance — Amazon’s finance teams are quietly automating complex tasks with AI agents. Here’s what that means.

 

1. Meta’s “AI glasses” debut stumbles in live demo 😎

At a late MetaConnect 2025 keynote, Mark Zuckerberg introduced the AI-powered Meta Ray-Ban Display arriving Sep. 30 for $799 and touted as the company’s first high-resolution AI glasses, plus refreshed Ray-Ban frames and a sports-focused Meta Vanguard.

According to Mashable Light Speed, live demos faltered as the LiveAI cooking assistant failed to follow basic prompts and a WhatsApp video call from CTO Andrew Bosworth repeatedly hijacked Zuckerberg’s view while the new Neural Band struggled to respond. The hardware headline is a bright 5,000-nit display and a wrist-worn Neural Band that lets you “handwrite” text in the air at roughly 30 words per minute, but core reliability questions remain after the onstage hiccups.

2. Notion unveils its first AI agent at “Make with Notion” �*

Notion launched an AI agent that can read across your pages and databases to auto-generate notes, competitor analyses, and feedback-ready landing pages, plus create or update pages, properties, and views. The agent runs multi-step tasks for up to 20 minutes across hundreds of pages, pulls data from linked tools like Slack, email, and Google Drive, and lets users set a profile page with style rules, source citation preferences, and editable memories.

Notion says scheduling and trigger-based custom agents, along with a template library, are coming soon.

3. Google bakes Gemini into Chrome across devices 🖥️

Google is rolling out agentic Gemini features inside Chrome on Mac, Windows and mobile, letting users summarize pages, work across tabs and trigger actions like scheduling or finding videos without leaving a tab.

The move arrives as OpenAI, Anthropic and Perplexity push their own agentic browser agents, signaling that the browser is becoming the frontline for consumer AI and a key distribution channel. Google says Gemini in Chrome will soon tie deeper into Calendar, YouTube, Maps and Workspace with enterprise-grade protections, plus new agentic tools that can book appointments and order groceries.

4. Nvidia sinks $5B in rival Intel as AI-PC alliance reshapes chip race 💾

Nvidia is investing $5 billion in rival Intel and partnering on custom data center and PC chips that blend Nvidia AI technology with Intel’s manufacturing and design footprint, even as the two remain competitors.

The timing matters because it follows a U.S. government move to take a 10% non-voting stake in Intel, signaling a coordinated push to secure domestic chipmaking and diversify away from single-foundry dependence. For the industry, this is classic coopetition that could ease GPU supply constraints, widen CPU-GPU integration paths and give Nvidia leverage beyond TSMC while giving Intel a credible route back into AI infrastructure.

5. New Pew Poll Finds Americans Sour on AI 🍋‍🟩

According to Pew Research Center, a June survey of just over 5,000 U.S. adults shows a sharp turn toward skepticism, with half now more concerned than excited about AI’s spread and majorities expecting harm to creativity, relationships and problem-solving.

The poll lands at a time when AI-generated content is increasingly photorealistic and text tools still hallucinate, while 76 percent say it is very important to know what was made by AI yet most doubt they can tell. A new study in the Journal of Marketing adds context, finding people with lower AI literacy tend to be more receptive, suggesting rising familiarity may be deflating the early mystique.

🦾How You Can Leverage:

Everything’s an AI agent now. 

That small feature in that one app has been rebranded an agent…. What? 

We get it. Every week brings another "revolutionary" AI agent that feels exactly like the last 7 you tried.

The $7.5 billion agent market just crossed the tipping point. 80 percent of enterprises are deploying them right now. Like, for real work. Every single major AI company shipped agent platforms in the past year.

This isn't hype anymore.

Microsoft, OpenAI, Google, Anthropic, and Amazon all have production-ready agent systems. The companies mastering these tools first are building sustainable competitive advantages while everyone else gets lost in comparison paralysis.

So today, we went deep on AI agents. (You asked us to! Lolz )

We just cut through hundreds of options to identify the 10 agents actually worth your time.

Let’s get it.  

1 –  Why 2025 Became The Agent Breakthrough Year 🚀

Two years ago, we said agents weren't ready.

Everything changed in the past 6-9 months.

Agents now show up directly inside the enterprise systems where you actually work. Your documents, emails, CRM, browsers, and development environments. No more duct tape integrations or hoping your custom setup holds together.

The underlying models got exponentially more capable. Most agents run five to ten different models simultaneously with sophisticated routing logic. A main agent breaks complex tasks into specialized sub-agents running coding models, reasoning models, and analysis engines.

Here's the breakthrough moment.

These models now plan their own routes, fix their own mistakes, and work autonomously for hours without human intervention. Replit Agent 3 runs for three full hours building complete applications. Claude Code works for 90 minutes straight on complex development tasks.

Try This: Test your current AI tools with this simple benchmark. Give them a task requiring 45 minutes of manual work and completely disconnect.

Come back in two hours. Real agents leave finished deliverables with step-by-step audit trails.

Everything else leaves error messages or half-completed work requiring human finishing touches.

2 – Hidden Costs That Destroy ROI 🔥

Agents can bankrupt unprepared teams.

Most platforms charge by usage, not subscription. When agents get stuck in endless loops, your API bill explodes while delivering zero business value.

We found agents accidentally updating hundreds of customer records with incorrect data. Others deploying broken code changes across entire applications. Some sending automated emails to prospects with completely fabricated information.

Smart companies implement approval workflows and rollback capabilities before granting write permissions. They treat agents like employees with full identity management and audit trails tracking every autonomous decision.

Microsoft Copilot Studio assigns Azure Entra IDs to each agent. IT departments can monitor agent actions across organizational systems just like tracking human employee access.

Try This: Before deploying any agent with write permissions, create a disaster recovery plan.

Document exactly how you'll identify agent errors quickly. Test your rollback procedures with sample data.

Calculate the true cost of autonomous mistakes including customer impact, data cleanup, and reputation damage.

3 – The 10 Agents Actually Worth Using ⚡

Most "agents" are rebranded chatbots with marketing budgets.

These ten platforms demonstrate true autonomous execution with persistent runtime capabilities.

ChatGPT Agent Mode runs scheduled tasks daily without human input. Easiest learning curve for understanding agent behavior. Microsoft Copilot Studio provides enterprise governance with approval workflows and identity management for organization-wide deployment.

Claude Code handles 90-minute autonomous coding sessions with sub-agent architecture for complex development workflows. AWS Bedrock Agents offers modular frameworks for custom enterprise builds with security controls.

Salesforce AgentForce automates CRM workflows for revenue teams already living in their ecosystem. Google Project Mariner runs 10 concurrent browser sessions for parallel web task execution.

Replit Agent 3 builds complete applications from plain English with three-hour autonomous runtime. Zapier Agents coordinate actions across 7,000+ enterprise applications without code.

GenSpark Super Agent produces research pages and presentations using nine different language models with specialized routing logic. Manus AI maintains persistent cloud sessions continuing work even when you disconnect for days.

Try This: Pick your most expensive manual workflow and test it on exactly two platforms from this list.

Don't compare five different agents across one use case. Choose two platforms, run five real test cases, and measure completion rates against human baseline performance.

Calculate true ROI including setup time, error correction, and ongoing governance overhead.

 🚨 Bonus Materials 🚨

Anotha one?!

We did it again.

If you wanna go deeper on our hand-vetted best AI agents out there, this guide is for you. To get access, just go repost today’s LinkedIn show and we’ll shoot it over.

(Yeah… we added 10 more agents and a WHOLE lot more to this list)

Reply

or to participate.