- Everyday AI
- Posts
- Perplexity Deep Research: What it is and if you should use it
Perplexity Deep Research: What it is and if you should use it
xAI releases Grok 3, Google boosts Chrome security with AI, OpenAI’s plans to avoid takeovers and more!
Everyday AI
February 18, 2025
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Outsmart The Future
Sup y’all
We gave Perplexity’s Deep Research tool a deep test today. Worth checking out.
Tomorrow, we’re doing a Deep Research Throwdown: Perplexity vs. OpenAI vs. Google. It’s a can’t miss.
(Side note – we’d include Grok’s new Deep Search tool, but it’s not yet available as far as we can tell.)
Jordan
Today in Everyday AI
7 minute read
🎙 Daily Podcast Episode: What is Perplexity Deep Research and how does it compare to the other LLMs’ Deep Research? We dive in. Give it a listen.
🕵️♂️ Fresh Finds: Meta’s new AI event, Topaz Labs’ AI diffusion model and Texas proposes new AI bill. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: xAI releases Grok 3, Google boosts Chrome security with AI and OpenAI’s plans to avoid takeovers. For that and more, read on for Byte Sized News.
🚀 AI In 5: Is Anthropic Claude’s Projects feature better than ChatGPT’s custom GPT? We put them to the test and you might be surprised by the results! See it here
🧠 Learn & Leveraging AI: We break down everything you need to know about Perplexity’s new Deep Research. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We talked about the EU launching an open source LLM project, South Korea banning DeepSeek downloads and the NYT unveiling new AI tools. Check it here!
Perplexity Deep Research: What it is and if you should use it 🧠
It's free.
It's fast.
But is it good?
Perplexity joined the Deep Research train, so we are giving it a thorough rundown. Is this your next AI sidekick?
Join the conversation and ask Jordan questions on Perplexity here.
Also on the pod today:
• Deep Research vs. competition 🥊
• Mechanics of Perplexity Deep Research ⚙️
• Results and Analysis of Deep Research 📊
It’ll be worth your 52 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Upcoming Everyday AI Livestreams
Wednesday, February 19th at 7:30 am CST ⬇️
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Motiff is an AI-powered UI design tool, Summer is an AI summary for your blog and WebSim is an AI web editor and simulator.
Meta – Meta has announced its new event called LlamaCon which will focus on its AI.
Trending in AI – AI tech’s role in warfare is being questioned after US tech giants supplied Israel with AI models.
AI Models – Topaz Labs is now letting users try its Project Starlight diffusion AI model.
AI Governance - Texas is proposing the Texas Responsible AI Governance Act (TRAIGA), a bill aimed to implement comprehensive AI regulations with a risk-based approach.
Google - Google Cloud has released a report on AI and cybercrime along with security best practices.
AI in Media – VFX house DNEG has unveiled a deal to buy AI platform Metaphysic.
Business of AI - Mastercard and Feedzai are teaming up to fight AI-powered scams.
AI Video - Pika Labs has released its official app on iOS.
1. xAI Releases Grok 3 🚀
Elon Musk's xAI has unveiled Grok 3, an AI model designed to compete with giants like OpenAI's GPT-4o and Google's Gemini, as reported by xAI. The ambitious launch, backed by an enormous data center in Memphis with around 200,000 GPUs, claims Grok 3 offers ten times the computing power of its predecessor, Grok 2.
The new model excels in handling complex math, science, and programming queries and introduces features like reasoning models and DeepSearch for enhanced information analysis and research capabilities. As xAI rolls out these innovations with exclusive access for X’s Premium+ subscribers, Musk also plans to open-source past versions.
2. OpenAI Shields from Takeovers with Special Board Rights 🛡️
In a strategic maneuver to prevent hostile takeovers, OpenAI is exploring the option of granting its non-profit board special voting rights, according to a report by the Financial Times. This comes on the heels of a massive $97.4 billion acquisition attempt led by Elon Musk, which the board swiftly rejected.
As OpenAI transitions from its "capped-profit" structure to a public benefit corporation, these rights could empower the board to override decisions by major investors, thereby preserving the company’s mission-driven focus.
3. AI Boosts Chrome's Defense 🔒
Google is rolling out an AI-powered "Enhanced Protection" feature in Chrome, offering real-time shields against malicious sites, downloads, and extensions. After three months in the experimental Chrome Canary, this update strengthens Safe Browsing by using AI to dynamically analyze and thwart threats, as reported by BleepingComputer.
Unlike previous methods, this AI-driven approach detects dangers even before they're flagged by Google.
4. Baidu Stumbles in AI Race Despite Revenue Uptick 📉
Baidu reported a smaller-than-expected 2% revenue drop for the last quarter, with a notable 26% jump in cloud revenue. However, the Chinese tech giant faces challenges as it loses ground to social apps like Xiaohongshu and Douyin in search and to startups like DeepSeek in AI.
Despite integrating DeepSeek's AI model into its chatbot, Baidu's shares fell over 4% in pre-market trading as investors remain wary of its weakening core business. With its leadership in AI at risk, Baidu's strategy shift to open-source models highlights the intense competition in China's tech industry.
5. Replit and AI Partners Transform Zillow's Workforce Into Coders 🧑💻️
Zillow's non-tech employees have morphed into software developers, thanks to Replit's partnership with Anthropic and Google Cloud. According to VentureBeat, this collaboration has empowered over 100,000 home applications on Google Cloud Run, addressing the looming developer shortage. Claude AI, Anthropic’s brainchild, shines by generating clean code and tackling complex problems strategically.
This initiative not only democratizes software development but also reshapes how businesses craft custom solutions, potentially revolutionizing the global software market projected to hit $700 billion by 2028.
6. Fluidstack Teams Up with France for AI Supercomputer 🇫🇷
In a move announced at the AI Action Summit in Paris, Fluidstack sealed a deal with the French government to build a massive AI supercomputer, slated to go live in 2026. The $10.4 billion project, powered by France's carbon-free nuclear energy, will host nearly 500,000 next-gen AI chips, making it one of the largest low-carbon AI supercomputers globally.
This ambitious initiative not only strengthens France's position in AI but also promises thousands of high-skilled jobs, providing a significant boost to both digital and energy sovereignty.
Anthropic Claude Projects: Better than ChatGPT’s GPTs?
Anthropic’s projects feature is pretty dang good!
The feature essentially allows you to train your chats.
So we were curious to see, how does this Projects feature stack up against ChatGPT’s custom GPTs?
We put the two to the test.
Check out today's AI in 5.
🦾How You Can Leverage:
When it comes to AI and LLMs, we all want fast and free, right?
So that’s why Perplexity’s new Deep Research mode is drool-worthy at first glance.
Perplexity just dropped their newest feature, a Deep Research mode that scours the web even more so than its normal modes do.
And it’s free.
So how does Perplexity’s new feature work?
And should you or your company be using it?
We went all-in on Perplexity’s new Deep Research tool on today’s show.
What did we find?
Let’s explore, shorties.
1 – The Secret Sauce Behind Perplexity’s Research Engine 🤫
While Google's deep research plays by the rules, Perplexity's version is that one detective who throws away the handbook.
And sometimes strikes gold.
Watch what happened during our Nike analysis. Ten initial searches. Standard stuff.
Then – plot twist. The AI noticed gaps in its financial data and went rogue.
Spawned new research paths. Pulled up specialized finance modes. Started cross-referencing Yahoo Finance with CNBC like a caffeinated day trader.
That's not a search engine. That's an investigation machine.
When we pushed it on the latest on DeepSeek, it devoured 55 sources in minutes. Price points. Regulatory frameworks. Competitive dynamics. Lightning fast. Almost too fast.
Try This:
Don't just run queries – study how this thing thinks.
Pull up the reasoning steps for each search.
Watch how it pivots when it hits dead ends. Notice which sources trigger new research paths. Then run the same query again, but this time, preempt its next move. Add specific parameters that guide it toward the sources you know it should be checking.
Think of it like teaching a brilliant but chaotic detective how to solve cases your way. Each iteration should be sharper than the last. Keep a log of which reasoning patterns produce the most accurate results. This isn't just about getting answers y’all.
It's about learning how to take what AI gives you and make it (and yourself) better.
2 – When Deep Research Goes Full Fiction Writer 😬
We tested Perplexity on our own company. Told it to do a creepy level of research on Jordan Wilson of Everyday AI.
Because what could go wrong?
No lolz.
First run: 243 sources. Including Reddit threads about NFL quarterback Russell Wilson and new dad advice.
Um?
But then it got creative. The AI invented an entire "human amplification principle" framework. Complete with implementation stages. Cited real sources that don't contain any of that information.
The scary part?
When we reran the same query two days apart, new hallucinations. Different fake metrics. Same confident tone.
Yikes.
Try This:
Create a hallucination detection system that would make a fact-checker weep with joy.
Run a deep research query on something you know cold.
Document every claim: True, False, Uncertain.
Rerun the same query 24 hours later. Watch for patterns in the lies.
Does it consistently make up metrics? Client stories? Implementation frameworks?
This isn't just verification – it's understanding how AI fabricates so you can spot it in unfamiliar territory.
3 – The Free vs. Real That Could Wreck Your Research 😳
The faster Perplexity spits out results, the more it can start writing creative fiction.
Remember that gorgeous Nike analysis chart? Clean metrics. Professional formatting. Zero source citations for the actual numbers.
Great.
Here's the real problem: Each new research path is another opportunity for hallucination.
In the ‘do a creepy in-depth search on Jordan Wilson and Everyday AI’ test, it cited our Chicago AI Week panel for statistics from that page that didn’t even exist.
The audacity is almost impressive.
Try This:
Use those five free daily searches strategically.
First search: Cast a wide net. Study the reasoning patterns.
Where did it look first? What triggered new search paths? What sources did it trust?
Second search: Same query, but guide its reasoning based on what you learned. Add specific parameters that steer it toward reliable sources.
Keep refining until you've got it trained to think like your best researcher. Remember: The goal isn't to copy-paste results – it's to understand how this thing thinks so you can make it think better.
This isn't just about using a new tool. It's about understanding how AI reasons, fails, and occasionally strikes gold. Because in the end, the best defense against hallucination is knowing exactly how it happens.
So, what’s our takeaway?
Don’t trust our limited testing. You’ve gotta put Perplexity’s Deep Research through the paces on your own.
Like we said, start with something that you know better than anyone, and ask for an insanely in-depth report.
For now, Perplexity’s Deep Research tool seems like a freebie lover’s dream.
(If you can avoid the hallucinations.)
Now This …
Let us know your thoughts!
Vote to see live results
Do you attend our livestreams?Every weekday, we bring you fresh AI insights, exclusive interviews, and breaking news with our Everyday AI livestream. |
Reply