• Everyday AI
  • Posts
  • OpenAI’s Deep Research: How it works and what to use it for

OpenAI’s Deep Research: How it works and what to use it for

DeepSeek's training cost revealed, OpenAI’s partnership in South Korea, Salesforce cuts 1,000 jobs, U.S. Gov. agency proposes AI strategy and more!

Author

Everyday AI
February 04, 2025

Outsmart The Future

Sup y’all 👋

Today’s show was a bit spicy. And for good reason. 

OpenAI’s Deep Research isn’t just an AI-powered answers engine. 

It’s an agentic step toward AGI. And it’s super impressive. 

Make sure you check out today’s livestream, and repost it if you want our step-by-step guide on business use cases for Deep Research. 

Trust us. You’ll want this free guide. 

✌️
Jordan 

(Let’s connect on LinkedIn. Just tell me you’re from the newsletter because Stranger Danger.) 

Today in Everyday AI
7 minute read

🎙 Daily Podcast Episode: OpenAI just released its new agentic tool, Deep Research. But wait..didn’t Google just release that? We dive into OpenAI’s new release and break down what you need to know. Give it a listen.

🕵️‍♂️ Fresh Finds: DeepSeek paves way for open-source AI, new proposed California AI bill and Snapchat’s new text-to-image model. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: DeepSeek’s true training cost revealed, OpenAI’s partnership in South Korea, Salesforce cuts 1,000 jobs and U.S. Gov. agency proposes AI strategy. For that and more, read on for Byte Sized News.

🚀 AI In 5: Want the cliff notes on how ChatGPT’s Task mode works? We break it down in this short review. See it here

🧠 Learn & Leveraging AI: We break down how Deep Research works and what OpenAI’s new agentic tool means for the AI landscape. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about OpenAI unveiling Deep Research, EU beginning AI law enforcement and SoftBank and OpenAI creating a Japanese venture. Check it here!

 OpenAI’s Deep Research - How it works and what to use it for 🔍

Another "Deep Research"?

OpenAI just released its version of Deep Research, not to be confused with Google's own Deep Research.

But not all AI-powered Deep Research tools are the same, believe it or not.

We show you how OpenAI's newest agentic tool works and how you can use it.

Join the conversation and ask Jordan questions on AI here.

Also on the pod today:

• OpenAI vs. Google Deep Research ⚔️
• Deep Research Use Cases 💼
• How Deep Research Works 🕵

It’ll be worth your 56 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 – Latitude is an open-source prompt engineering platform, Height.app is an autonomous project management tool and Handinger helps extract data from the internet.

Trending in AI – DeepSeek’s success is providing a runway for open-source models like Meta’s Llama.

AI Governance – A proposed California bill would require AI companies to remind kids that chatbots aren’t real people.

Adobe - Adobe Acrobat’s AI chatbot can now understand contract jargon.

AI Models – 
Snapchat has unveiled a new text-to-image model for mobile devices.

A new EU alliance called OpenEuroLLM has emerged, to develop foundation models for the EU.

AI in Education - California State University is integrating ChatGPT for half a million students.

Read This - MIT has launched the Generative AI Impact Consortium, joining forces with industry giants like OpenAI and Coca-Cola.

AI Tech – Verizon is bundling Google AI into its wireless plan.

AI Agents – LogicStar is building AI agents for app maintenance.

Money in AI - Neuralk-AI has raised $4M to transform tabular data with AI

1. DeepSeek’s Low-Cost Training is Reportedly False 🤯

Chinese startup DeepSeek claims to have trained a cutting-edge AI model, R1, with surprisingly low resources. However, industry analyst SemiAnalysis reports that the company has actually invested $1.6 billion in hardware, contradicting their claims of frugal innovation.

DeepSeek's approach of hiring exclusively from within China and running its own data centers has allowed it to maintain rapid development cycles. This revelation underscores the complexity of AI advancements, reminding us that significant financial investments remain crucial for staying competitive in the AI industry.

2. OpenAI Expands Asia Presence with Kakao Partnership 🤝

In a strategic move, OpenAI has joined forces with South Korea's Kakao, marking a significant expansion into Asian markets amidst growing competition from Chinese AI firm DeepSeek. Announced at a Seoul event, the partnership will see the development of a Korean-language AI assistant, Kanana, and integration of OpenAI tech into KakaoTalk, one of the region’s most popular messaging apps.

Meanwhile, SoftBank's $3 billion commitment to OpenAI further solidifies its presence in Japan, aiming to deploy AI solutions across its subsidiaries through a joint venture.

3. Scale AI CEO Heads to DC to Discuss AI Alarm Bells 🚨

Scale AI's CEO Alexandr Wang is in Washington, D.C., meeting with lawmakers and Trump administration officials to discuss the growing AI threat from China. Wang's visit underscores the urgency of the AI race, especially after China's DeepSeek demonstrated its capability to match U.S. advancements with minimal resources.

This comes as American tech leaders, including OpenAI's Sam Altman, also engage Washington to highlight AI's potential for economic growth and national security.

4. AI-First Strategy Unveiled For U.S. Government 🇺🇸

Thomas Shedd, a former Tesla engineer now at the helm of Technology Transformation Services, has reportedly unveiled an "AI-first strategy" for the U.S. government. Emphasizing a startup-like approach, Shedd's plan includes creating a centralized database for AI analysis amidst looming budget cuts of up to 50%.

This strategy reflects Silicon Valley's growing influence within government circles under the Trump administration, blending tech innovation with fiscal restraint.

5. Salesforce Cuts Jobs Amid AI Expansion ✂️

Salesforce is reportedly slashing over 1,000 jobs while recruiting new staff to focus on AI sales, as per Bloomberg. The company, known for its customer management software, had nearly 73,000 employees at the end of its last fiscal year.

Despite layoffs, Salesforce is honing in on profit margins under activist investor pressure, recently launching a generative AI-powered point-of-sale system to transform shopping experiences.

6. Deepfake Alert: ByteDance's OmniHuman-1 Breaks New Ground 😳

ByteDance's cutting-edge AI system, OmniHuman-1, has showcased its ability to create near-flawless deepfake videos, as reported by TechCrunch's Kyle Wiggers. This breakthrough in AI technology only requires a single reference image and audio to produce convincing, adjustable-length clips, raising the stakes in the world of digital impersonation.

Despite the impressive advancements, the system struggles with low-quality images and certain poses, highlighting ongoing challenges.

7. VP Vance Ventures to Paris for AI Talks 🇫🇷

U.S. Vice President JD Vance is set to dive into the global AI conversation at the AI Action Summit in Paris next week, marking his debut on the international stage since taking office, according to the Associated Press.

This summit, co-hosted by French President Emmanuel Macron and Indian Prime Minister Narendra Modi, will see top global leaders, including China's Vice Premier Ding Xuexiang, gather to navigate AI's rapid evolution, which has rattled the tech industry.

ChatGPT Tasks Review - Agentic AI or small feature update?

With all these AI announcements over the past few weeks, you may have missed OpenAI’s ChatGPT Tasks.

We’re going over what they are, how they work and the pros and cons.

Plus we’ll talk about advanced ways this new feature can help automate your work.

🦾How You Can Leverage:

No one caught the real story in OpenAI's latest announcement for its agentic researcher, Deep Research. 

While everyone's testing their shiny new research tool, we spotted something way bigger: They're building AI that discovers new knowledge for itself.

Not for us. For ITSELF.

So is OpenAI’s Deep Research a new tool/mode that changes how knowledge work can be done? 

Yes. 

Is it also a sneaky powerful step in OpenAI’s larger mission to get the training data it needs to achieve its AGI mission? 

Also yes. 

Today on Everyday AI, we broke down the big picture, and step-by-step process, to use this new juggernaut of a tool. 

(And yes, it’s VERY different from Google’s Deep Research tool, which we LOVE!) 

The good news? 

OpenAI’s Deep Research is a legit gold mine. 

The bad news?

Right now, it’s only available on OpenAI’s pricier $200/mo ChatGPT Pro plan. (But, OpenAI said it’ll roll out to normal ChatGPT Plus and free users soon.) 

With the whole, ‘you probably can’t access this now’ thing outta the way, here’s the 1-2-3 of what you need to know.

1 – OpenAI’s Deep Research: Agentic and Omnidirectional 🧠

Forget everything you know about AI research tools. 

This isn't Google's cached page crawler wearing a fancy hat. (Though, again, we’re BIG fans of Google’s variety of Deep Research.) 

OpenAI’s Deep Research thinks like a seasoned investigative journalist. It starts with your query, sure. 

But when it spots something interesting? 

It pivots. FAST.

Example: When analyzing Nike's financials on today’s livestream, it didn't just grab the numbers and bounce. It noticed discrepancies in gross margin reporting across different sources. 

Then it automatically launched a new investigation path. Zero human input needed.

That's not web browsing. That's legitimate management consultant work. (More on that below) 

The receipts are in the Activity tab. You can watch in real-time as it discovers new leads, follows promising threads, and builds connections that would take human researchers days to piece together.

Try This:

You didn’t repost today’s livestream yet? 

Awkward. 

You might wanna do that, and we’ll share our Deep Research Use Cases with ya! 

2 – The Hidden Power of o3’s Reasoning Engine 💪

Let's talk about that 26.6% score on humanity's last exam. IYKYK.

For everyone else: This is a new benchmark for testing AI capabilities. 

GPT-4o scored 3.3%. 

Claude? 4.3%. 

Gemini Thinking? 6.2%

Deep Research: 26.6%

This massive performance gap isn't just about processing power. 

It's about how deep research actually thinks. 

The underlying o3 model powering Deep Research isn't just scanning text - it's building complex mental models of how information connects.

And with that 200,000 token context window? It can juggle more information simultaneously than an entire research department. 

We're talking entire financial reports, market analyses, and competitive intelligence. 

All at once.

Try This:

Leverage that massive context window like a pro. Feed it your highest-value documents first - earnings calls, SEC filings, internal reports. Then let it loose on the web to find supporting evidence and contrasting viewpoints.

Pro tip: Create source tiers. Rank your industry's most reliable publications and data sources. Then teach deep research which ones deserve priority treatment. The quality of your output skyrockets when you guide its initial direction.

3 – Why Every Management Consultant Just Started Sweating 😰

The math is brutal. Absolutely brutal.

Traditional market research: 200+ billable hours, $50,000+, weeks of waiting.

Deep research: 30 minutes, fully cited report, $200/month subscription.

Things are about to get awkward. 

This isn't just disruption - it's an extinction-level event for old-school consulting.

(Sorry Big 4 peeps. Using your own GPT- or 03-powered model with a RAG pipeline isn’t enough. You’ve gotta be using Deep Research, among other tools.) 

The most devastating part? The output quality is often BETTER than traditional consulting work. 

No human biases. No recycled frameworks. No last year's slides with updated logos.

Just pure, data-driven insights delivered faster than you can expense lunch.

Try This:

Next time a consulting firm quotes you $50k for market research, run this playbook:

Set up parallel research streams. Have deep research tackle the same questions as your consultants. Document everything - sources, methodologies, time spent, insights generated.

Then comes the fun part. Lolz. 

Start asking questions about their research process. How are they using AI tools? What percentage of their work is now automated? Why aren't those efficiency gains reflected in their pricing?

(Then it might get awkward when you show them YOUR research findings vs. theirs. Rut-roh.) 

Remember: They're (hopefully) already using these tools. 

The only question is whether you're getting the cost savings, or if they’re pocketing it. 

Or paying premium prices for AI-generated insights with a fancy letterhead.

⌚

Numbers to watch

$235 Million

Canada’s StackAdapt has raised $235M for its AI-based programmatic platform.

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.

Login or Subscribe to participate in polls.

Reply

or to participate.